FORM 2 THE PATENTS ACT, 1970 (39 of 1970) & THE PATENTS RULES, 2003 COMPLETE SPECIFICATION [See section 10, Rule 13] RADIO COMMUNICATION DEVICE AND SIGNAL DIVISION METHOD; PANASONIC CORPORATION, A CORPORATION ORGANIZED AND EXISTING UNDER THE LAW'S OF JAPAN, WHOSE ADDRESS IS 1006. OAZA KADOMA, KADOMA-SHI, OSAKA 5718501, JAPAN THE FOLLOWING SPECIFICATION PARTICULARLY DESCRIBES THE INVENTION AND THE MANNER IN WHICH IT IS TO BE PERFORMED. DESCRIPTION Technical Field The present invention relates to a radio communication apparatus and a signal division method. Background Art In 3GPP LTE (3rd Generation Partnership Project Long Term Evolution), active studies are underway on standardization of a mobile communication standard to realize low-delay and high-speed transmission, To realize low-delay and high-speed transmission, OFDM (Orthogonal Frequency Division Multiplexing) is adopted as a downlink (DL) multiple access scheme and SC-FDMA (Single-Carrier Frequency Division Multiple Access) using DFT (Discrete Fourier Transform) precoding is adopted as an uplink (UL) multiple access scheme. SC-FDMA using DFT precoding uses a DFT matrix (precoding matrix or DFT sequence) represented by, for example, an NxN matrix. Here, "N is the size of DFT (the number of DFT points). Furthermore, in an NxN DFT matrix, N (Nxl) column vectors are orthogonal to each other in DFT size N. SC-FDMA using DFT precoding forms an SC-FDMA signal (spectrum) by spreading and code-multiplexing a symbol sequence using this DFT matrix. Furthermore, standardization of LTE-Advanced (or IMT (International Mobile Telecommunication)-Advanced) to realize higher-speed communication than LTE has started. In LTE-Advanced, a 2 radio communication base station apparatus (hereinafter referred to as "base station") and a radio communication terminal apparatus (hereinafter referred to as "terminal") which are communicable using a wideband of, for example, 40 MHz or higher are expected to be introduced to realize higher-speed communication. As for an LTE uplink, uplink frequency resource allocation is limited to such allocation that SC-FDMA signals are mapped to continuous frequency bands in a localized manner to maintain single-carrier characteristics (e.g. low PAPR (Peak-to-Average Power Ratio) characteristics) of a transmission signal for realizing high coverage. However, when frequency resource allocation is limited as described above, vacancy is produced in uplink shared frequency resources (e.g. PUSCH (Physical Uplink Shared CHannel)) and the efficiency of the use of frequency resources becomes worse. Thus, as a prior art for improving the efficiency of the use of frequency resources, clustered SC-FDMA (C-SC-FDMA) is proposed which divides an SC-FDMA signal into a plurality of clusters and maps the plurality of clusters to discontinuous frequency resources (e.g. see non-patent literature 1). In C-SC-FDMA of the above prior art, a terminal generates C-SC-FDMA signals by dividing an SC-FDMA signal (spectrum) generated through DFT processing into a plurality of clusters. The terminal then maps the plurality of clusters to discontinuous frequency resources (subcarriers or resource blocks (RB)). On the other hand, a base station applies frequency domain equalization (FDE) processing to the received C-SC-FDMA signals (plurality of clusters) and combines the plurality of clusters after the equalization. The base station then applies IDFT (Inverse Discrete Fourier Transform) processing to the combined signal and thereby obtains a time domain signal. C-SC-FDMA can allocate frequency resources among a plurality of terminals more flexibly than SC-FDMA by mapping the plurality of clusters to a plurality of discontinuous frequency resources, and can thereby improve the efficiency of the use of frequency resources and multiuser diversity effect. Furthermore, C-SC-FDMA has a smaller PAPR than that of OFDMA (Orthogonal Frequency Division Multiple Access), and can thereby expand uplink coverage more than OFDMA. Furthermore, a C-SC-FDMA configuration can be easily realized by only adding a component that divides an SC-FDMA signal (spectrum) into a plurality of clusters to the terminal and adding a component that combines a plurality of clusters to the base station in the conventional SC-FDMA configuration. Citation List Non-Patent Literature NPL 1 Rl-081842, "LTE-A Proposals for evolution," 3GPP RAN WG1 #53, Kansas City, MO, USA, May 5-9, 2008 Summary of Invention Technical Problem According to the above prior art, the base station divides an SC-FDMA signal (spectrum) of each terminal with an arbitrary frequency according to a state of availability of uplink frequency resources and a condition of the propagation path between a plurality of terminals and the base station, allocates a plurality of clusters thereby generated to a plurality of uplink frequency resources respectively and reports information showing the allocation result to the terminals. The terminal divides the SC-FDMA signal (spectrum) which is the output of DFT processing with an arbitrary bandwidth, maps the plurality of clusters to a plurality of uplink frequency resources allocated by the base station respectively and thereby generates C-SC-FDMA signals. However, since a wide uplink radio frequency band (wideband radio channel) is frequency selective, the frequency correlation between channels through which a plurality of clusters mapped to different discontinuous frequency bands propagate decreases. Thus, even when the base station equalizes C-SC-FDMA signals (a plurality of clusters) through FDE processing, the equalization channel gain (that is, frequency channel gain after FDE weight multiplication) may considerably differ among the plurality of clusters. Therefore, the equalization channel gain may drastically change at a combining point (that is, the point of division at which the terminal divides the SC-FDMA signal) of the plurality of clusters. That is, a discontinuous point may occur in a variation (that is, envelope of reception spectrum) in the equalization channel gain at the combining point of the plurality of clusters. Here, to keep minimal the loss of orthogonality of a DFT matrix in all frequency bands (that is, the sum of frequency bands to which a plurality of clusters are mapped) to which C-SC-FDMA signals are mapped, the equalization channel gain in all frequency bands to which the plurality of clusters are mapped needs to be a slow variation. Thus. when a discontinuous point occurs in a variation of the equalization channel gain at a combining point of the plurality of clusters as in the above described prior art, the orthogonality of the DFT matrix is considerably destroyed in the frequency band to which the C-SC-FDMA signals are mapped. Therefore, the C-SC-FDMA signals are more impacted by inter-symbol interference (ISI) caused by the loss of orthogonality of the DFT matrix. Especially when high-level M-ary modulation such as 64 QAM whose Euclidian distance between signal points is very short is used, the C-SC-FDMA signals are more impacted by ISI, and therefore deterioration of transmission characteristics is greater. Furthermore, as the number of clusters (the number of fractions of SC-FDMA signal) increases, the number of discontinuous points between clusters increases, and therefore ISI caused by the loss of orthogonality of the DFT matrix further increases. The present invention has been implemented in view of such problems and it is therefore an object of the present invention to provide a radio communication apparatus and a signal division method capable of reducing ISI caused by the loss of orthogonality of a DFT matrix even when an SC-FDMA signal is divided into a plurality of clusters and the plurality of clusters are mapped to discontinuous frequency bands respectively, that is, when C-SC-FDMA is used. Solution to Problem A. radio communication apparatus of the present invention adopts a configuration including a conversion section that generates a frequency domain signal by applying DFT processing to a symbol sequence using a DFT matrix, a division section that divides the signal with a partially orthogonal bandwidth corresponding to a partially orthogonal vector length of some of a plurality of column vectors constituting the DFT matrix and generates a plurality of clusters a.nd a mapping section that maps the plurality of clusters to a plurality of discontinuous frequency bands respectively. A signal division method of the present invention divides a frequency domain signal with a partially orthogonal bandwidth corresponding to a partially orthogonal vector length of some of a plurality of column vectors constituting a DFT matrix used to convert a time domain symbol sequence to the frequency domain signal and generates a plurality of clusters. Advantageous Effects of Invention When dividing an SC-FDMA signal into a plurality of clusters and mapping the plurality of clusters to discontinuous frequency bands (when using C-SC-FDMA), the present invention can reduce IS1 caused by the loss of orthogonality of a DFT matrix. Brief Description of Drawings FIG. I is a block diagram of a terminal According to Embodiment 1 of the present invention; FIG.2 is a diagram showing DFT processing according to Embodiment 1 of the present invention; FIG.3 is a diagram showing an example of DFT matrix according to Embodiment 1 of the present invention; FIG.4A is a diagram showing a partially orthogonal relationship according to Embodiment 1 of the present invention (when |F| = 1); FIG.4B is a diagram showing a partially orthogonal relationship according to Embodiment 1 of the present invention (when |I|=2); FIG.4C is a diagram showing a partially orthogonal relationship according to Embodiment 1 of the present invention (when |I|=3); FIG.5A is a diagram showing division processing and mapping processing according to Embodiment 1 of the present invention; FIG.5B is a diagram showing a signal after FDE according to Embodiment 1 of the present invention; F1G.5C is a diagram showing a signal after combining according to Embodiment 1 of the present invention; FIG.6 is a diagram showing an orthogonal relationship of column vectors according to Embodiment 1 of the present invention; FIG.7 is a diagram showing an orthogonal relationship of column vectors according to Embodiment 1 of the present invention; F1G.8 is a diagram showing frequency interleaving processing according to Embodiment 1 of the present invention; F1G.9 is a block diagram of a terminal according to Embodiment 2 of the present invention; FIG.10A is a diagram showing precoding processing according to Embodiment 2 of the present invention; FIG.IOB is a diagram showing precoding processing according to Embodiment 2 of the present invention; FIG. 11 is a diagram showing processing using FSTD according to Embodiment 2 of the present invention; FIG.I2 is a diagram showing processing using FSTD according to Embodiment 3 of the present invention; FIG.13 is a diagram showing processing using FSTD according to Embodiment 3 of the present invention; FIG.14 is a diagram showing a relationship between a multiplier and a cluster size according to Embodiment 4 of the present invention; F1G.15 is a block diagram of a terminal according to Embodiment 5 of the present invention; FIG. 16 is a block diagram of a base station according to Embodiment 5 of the present invention; FIG.17A is a diagram showing shifting processing according to Embodiment 5 of the present invention (when z=0); FIG.17B is a diagram showing shifting processing according to Embodiment 5 of the present invention (when z=3); FIG. 18A is a diagram showing DFT output according to Embodiment 5 of the present invention; FIG.18B is a diagram showing shifting processing according to Embodiment 5 of the present invention; FIG. I8C is a diagram showing division processing and mapping processing according to Embodiment 5 of the present invention; FIG.19 is a block diagram of a terminal according to Embodiment 5 of the present invention; FIG.20 is a block diagram of a terminal according to Embodiment 6 of the present invention; FIG.21A is a diagram showing DFT output according to 9 Embodiment 6 of the present invention; FIG.21B is a diagram showing shifting processing according to Embodiment 6 of the present invention; FIG.21C is a diagram showing division processing and mapping processing according to Embodiment 6 of the present invention; FIG.22A is a diagram showing DFT output according to Embodiment 6 of the present invention; FIG.22B is a diagram showing shifting processing according to Embodiment 6 of the present invention; FIG.22C is a diagram showing division processing and mapping processing according to Embodiment 6 of the present invention; FIG.23 is a block diagram of a terminal according to Embodiment 7 of the present invention; FIG.24 is a diagram showing frequency shifting processing and space shifting processing according to Embodiment 7 of the present invention; FIG.25 is a diagram showing frequency shifting processing and space shifting processing according to Emh0diment 7 of the present invention; and F1G.26 is a diagram showing shifting processing according to Embodiment 8 of the present invention. Description of Embodiments Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. A case will be described below where a terminal provided with a radio communication apparatus according to the present invention transmits a C-SC-FDMA signal to a base station. (Embodiment 1) F1G.1 shows a configuration of terminal 100 according to the present embodiment. In terminal 100, radio receiving section 102 receives a control signal transmitted from a base station (not shown) via antenna 101, applies reception processing such as down-conversion and A/D conversion to the control signal and outputs the control signal subjected to the reception processing to demodulation section 103. This control signal includes frequency resource information showing uplink frequency resources allocated to each terminal and MCS information showing MCS (Modulation and channel Coding Scheme) set in each terminal. Demodulation section 103 demodulates the control signal and outputs the demodulated control signal to decoding section 104. Decoding section 104 decodes the control signal and outputs the decoded control signal to extraction section 105. Extraction section 105 extracts frequency resource information directed to terminal 100 included in the control signal inputted from decoding section 104 and outputs the extracted frequency resource information to control section 106. Control section 106 receives category information of the terminal including a DFT size (the number of DFT points) of a DFT matrix to be used in DFT section 110 and partially orthogonal condition information showing a partially orthogonal condition of a C-SC-FDMA signal as input and also receives frequency resource information reported from the base station from extraction section 105 as input. Control section 106 calculates the number of clusters generated by division section 111 by dividing an SC-FDMA signal (that is, the output of DFT section 110) and the cluster size showing a bandwidth of each cluster based on DFT size information (category information) showing the DFT size of the terminal, partially orthogonal condition information and frequency resource information reported from the base station. Suppose it is determined in advance between the base station and the terminal that when an SC-FDMA signal (spectrum) is divided into a plurality of clusters, the SC-FDMA signal (spectrum) is divided in order from a lower frequency portion of the spectrum (smaller output number of DFT section 110) or from a higher frequency portion of the spectrum (larger output number of DFT section 110). Control section 106 calculates frequency resources to which C-SC-FDMA signals (a plurality of clusters) of terminal 100 are mapped based on the calculated number of clusters and the cluster size. For example, control section 106 calculates frequency resources to which clusters are mapped in order from a cluster of a lower frequency (cluster with a smaller output number of DFT section 110) or a cluster of a higher frequency (cluster with a larger output number of DFT section 110) of the plurality of clusters generated through division. Control section 106 then inputs cluster information including the calculated number of clusters and cluster size to division section 111 and outputs mapping information showing frequency resources to which C-SC-FDMA signals (a plurality of clusters) of terminal 100 are mapped to mapping section 112. Coding section 107 encodes a transmission bit sequence and outputs the coded transmission bit sequence to modulation section 108. Modulation section 108 modulates the transmission bit sequence inputted from coding section 107 to generate a symbol sequence and outputs the symbol sequence generated to multiplexing section 109. Multiplexing section 109 multiplexes pilot signals and the symbol sequence inputted from modulation section 108. Multiplexing section 109 outputs the symbol sequence multiplexed with the pilot signals to DFT section 110. For example, a CAZAC (Constant Amplitude Zero Auto Correlation) sequence may be used as the pilot signals. Furthermore, although FIG. 1 adopts a configuration in which the pilot signals and the symbol sequence are multiplexed before applying DFT processing, a configuration in which the pilot signals and the symbol sequence are multiplexed after applying the DFT processing may also be adopted. DFT section 110 generates frequency domain signals (SC-FDMA signals) by applying DFT processing to the time domain symbol sequence inputted from multiplexing section 109 using a DFT matrix. DFT section 110 outputs the generated SC-FDMA signals (spectrum) to division section 111. Division section 111 divides the SC-FDMA signal (spectrum) inputted from the DFT section 110 into a plurality of clusters according to the number of clusters and the cluster size indicated in the cluster information inputted from control section 106. To be more specific, division section 111 generates a plurality of clusters by dividing the SC-FDMA signal (spectrum) with a bandwidth (partially orthogonal bandwidth) corresponding to a length (vector length) of some of the plurality of column vectors constituting the DFT matrix used in DFT section 110 and partially orthogonal to each other. Division section 11J then outputs C-SC-FDMA signals made up of the plurality of clusters generated to mapping section 112. Details of the method of dividing the SC-FDMA signal (spectrum) in division section 111 will be described later. Mapping section 112 maps the C-SC-FDMA signals (a plurality of clusters) inputted from division section 111 to frequency resources (subcarriers or RBs) based on mapping information inputted from control section 106. For example, mapping section 112 maps the plurality of clusters making up the C-SC-FDMA signals to a plurality of discontinuous frequency bands respectively. Mapping section 112 then outputs the C-SC-FDMA signals mapped to the frequency resources to IFFT section 113. IFFT section 11 3 generates a time-domain C-SC-FDMA signal by performing IFFT on the plurality of frequency bands inputted from mapping section 112 to which the C-SC-FDMA signals are mapped. Here, IFFT section 113 inserts 0's in frequency bands other than the plurality of frequency bands to which the C-SC-FDMA signals (plurality of clusters) are mapped. IFFT section 113 then outputs the time-domain C-SC-FDMA signal to CP (Cyclic Prefix) insertion section 114. CP insertion section 114 adds the same signal as that at the end of the C-SC-FDMA signal inputted from IFFT section 11 3 to the head of the C-SC-FDMA signal as a CP. Radio transmitting section 115 applies transmission processing such as D/A conversion, amplification and up-conversion to the C-SC-FDMA signal and transmits the signal subjected to the transmission processing to the base station via antenna 101, On the other hand, the base station performs FDE processing of multiplying the C-SC-FDMA signals (a plurality of clusters) transmitted from each terminal by an FDE weight and combines the C-SC-FDMA signals (the plurality of clusters) after the FDE processing in the frequency domain. The base station obtains a time domain signal by applying IDFT processing to the combined C-SC-FDMA signal. Furthermore, the base station generates channel quality information (e.g. CQI: Channel Quality Indicator) of each terminal by measuring an SINR (Signal-to-Interference plus Noise power Ratio) for each frequency band (e.g. subcarrier) between each terminal and the base station using pilot signals transmitted from each terminal. The base station then schedules allocation of uplink frequency resources (e.g. PUSCH) of each terminal using CQI and QoS (Quality of Service) or the like of a plurality of terminals. The base station then reports frequency resource information showing the uplink frequency resource allocation result (that is, the scheduling result) of each terminal to each terminal. For example, PF (Proportional Fairness) may be used as an algorithm used when the base station allocates frequency resources to each terminal. Furthermore, the base station controls the number of clusters and the cluster size using the DFT size and partially orthogonal condition as in the case of control section 106 of terminal 100 and combines the C-SC-FDMA signals (the plurality of clusters) based on the number of clusters and the cluster size. Next, details of the SC-FDMA signal (spectrum) division method by division section 111 will be described. Here, the SC-FDMA signal which is the output of DFT section 110 is configured by applying orthogonal frequency spreading to each symbol of a symbol sequence in a frequency band corresponding to the DFT size (column vector length) of the DFT matrix and code-multiplexing each symbol after the orthogonal frequency spreading. Here, assuming the DFT size is N, the DFT matrix used in DFT section 110 can be expressed by NxN matrix F=[f0, fi, ... , fN-1]- Furthermore, fi (i = 0 to N-l) is an Nxl column vector having (l/VN)exp(-j2;i(i*k)/N) (k=0 to N-l) as a k-th element. Furthermore, all column vectors fi (i=0 to N-l) are orthogonal to each other in DFT size N. That is, DFT section 110 multiplies M symbols (e.g. symbols #0 to #N-1) constituting the symbol sequence by respective column vectors fj (i=0 to N-l) of the DFT matrix, and thereby makes all symbols (symbols #0 to #N-I) orthogonal to each other in an orthogonal bandwidth (that is, bandwidth to which N symbols are mapped) corresponding to column vector length N. For example, in the case of DFT size N=8, a symbol sequence made up of eight symbols #0 to #7 as shown in the upper part of FIG.2 is inputted to DFT section 110. As shown in the lower part of FIG.2, DFT section 110 frequency-spreads symbols #0 to #7 with column vectors fo to f n of the DFT matrix respectively. DFT section 110 then code-multiplexes frequency-spread symbols #0 to #7. This allows an SC-FDMA signal having an orthogonal bandwidth corresponding to DFT size N to be obtained. Furthermore, FIG.3 shows an example of DFT matrix when DFT size N=8. That is, column vector fi (i=0 to 7) is an 8x1 column vector which has (l/V8)exp(-j2π(i*k)/8) as a k-th (where k=0 to 7) element. Furthermore, column vectors fo to f7 are orthogonal to each other in DFT size N = 8. Here, column vector fj of DFT matrix F is not only orthogonal to all other column vectors in DFT size N but also partially orthogonal to some other column vectors in vector length N' (where N' According to the present division method, division section 111 divides an SC-FDMA signal with partially orthogonal bandwidth B' (=N'*Bsub) corresponding to vector length N' calculated according to equation 1. In the following descriptions, suppose the number of clusters is 2, one cluster size is partially orthogonal bandwidth B' that satisfies equation 2 (or equation 1), and the other cluster size is differential bandwidth B"(=B-B') between orthogonal bandwidth B and partially orthogonal bandwidth B'. Furthermore, suppose DFT size N is 8. Thus, division section 111 divides the SC-FDMA signal (spectrum) inputted from DFT section 110 into two clusters; cluster #0 and cluster #1 as shown in FIG.5A. To be more specific, division section Ill divides the SC-FDMA signal having orthogonal bandwidth B with partially orthogonal bandwidth B' calculated according to equation 2. In other words, division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' calculated according to equation 1. Thus, division section 111 generates cluster #0 having partially orthogonal bandwidth B' and cluster #1 having bandwidth B" (=B-B') which is the difference between orthogonal bandwidth B and partially orthogonal bandwidth BAs shown in FIG.5A, mapping section 112 then maps cluster #0 and cluster #1 to two discontinuous frequency bands respectively. On the other hand, the base station receives a C-SC-FDMA signal made up of cluster #0 and cluster #1 shown in F1G.5A. The base station applies FDE processing to the C-SC-FDMA signal and thereby obtains a C-SC-FDMA signal after the FDE as shown in FIG.5B. The base station then combines cluster #0 and cluster #1 after the FDE shown in FIG.5B and thereby generates a signal having orthogonal bandwidth B (=B'+B") of the DFT matrix as shown in FIG.5C. As shown in FIG.5C, the variation of the equalization channel gain becomes discontinuous at a combining point between cluster #0 and cluster #1 . On the other hand, the variation of the equalization channel gain is slow in each cluster. Thus, ISI between multiplexed symbols corresponding to column vectors f; and fi that satisfy equation 2 or equation 1 (that is, between partially orthogonal multiplexed symbols) is reduced in cluster #0. Thus, in cluster #0 (that is, cluster having partially orthogonal bandwidth B'), it is possible to reduce iSl caused by a drastic variation of the equalization channel gain at the combining point (dividing point of the SC-FDMA signal) between cluster #0 and cluster Thus, according to the present division method, although a variation of the equalization channel gain becomes discontinuous at a combining point of a plurality of clusters, it is possible to reduce the loss of orthogonality between multiplexed symbols in a cluster having a partially orthogonal bandwidth. Therefore, according to the present division method, it is possible to reduce ISI caused by a drastic variation of the equalization channel gain even when the SC-FDMA signal is divided into a plurality of clusters. According to the present division method, division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' in which (|I|/|i-i'|)"' in equation I is 2 or more and less than N and at the same time one of divisors of N. This will be described more specifically below, Here, suppose DFT size N is 12 and the number of clusters is 2. When N=I2, divisors of N=12, which are 2 or more and less than 12, are 2, 3, 4 and 6. Thus, division section 111 selects one of (|I|/|i-i'()"'=2, 3, 4, 6 which is the reciprocal of (|I|/|i-i'|) shown in equation 1. That is, division section 111 selects one of vector lengths N'=6, 4, 3 and 2 according to equation I. That is, column vector fi and column vector fi' that satisfy (|I|/|i-i'|) = l/2, 1/3, 1/4 and 1/6 respectively in equation 1 are partially orthogonal in vector lengths N'=6, 4, 3 and 2 respectively. When, for example, dividing column vector fi (i=0 to 11) with vector length N'=6 (that is, when (|I|/|i-i'j)"'=2), division section 111 assumes vector length N' of cluster #0 to be 6 and assumes vector length N" of cluster #1 to be 6 (=N-N' = 12-6). That is, division section 111 divides the SC-FDMA signal having orthogonal bandwidth B (=N*BSub=12Bsub) into cluster #0 having partially orthogonal bandwidth B' (=N'*Bsub=6Bsub) and cluster #1 having bandwidth B" (=N"*Bsub=6Bsub). The same applies to cases where vector length N'=4, 3, 2. Thus, combination (N\ N") of vector lengths of two clusters (cluster #0 and cluster #1) including the cluster of vector length M' calculated using the present division method is one of (6, 6), (4, 8), (3, 9) and (2, 10). That is, all combinations of vector lengths of the two clusters are integers. Therefore, while the DFT size (the number of DFT points) of the DFT matrix takes an integer value of 0 to N-1, vector length N' and vector length N"=(N-N') that divide column vector fi can always be integer values without becoming fractions. In other words, partially orthogonal bandwidth B' that divides orthogonal bandwidth B(=N*BSUb) can always be limited to an integer multiple of Bsub. Thus, according to the present division method, it is possible to improve affinity between DFT processing of outputting an SC-FDMA signal using DFT size N, which is an integer value, and division processing of dividing the SC-FDMA signal, which is the output of the DFT processing, into a plurality of clusters while obtaining effects similar to those of division method 1. According to the present division method, division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N', which is a multiple of a prime number. This will be described more specifically below. For example, division section 111 assumes vector length N' to be multiple aoxo (where coefficient a0 is an integer equal to or greater than 1) of prime number x0. Here, suppose DFT size N is 12 and the number of clusters is 2. Furthermore, suppose prime number XO=3 and coefficient ao=3. Thus, division section 111 assumes vector length N' of cluster #0 to be 9 (=3x3) and vector length N" of cluster #1 to be 3 (=N-N' = 1 2-9). That is, division section 111 divides the SC-FDMA signal having orthogonal bandwidth B (=N*Bsub=l 2BSUb) corresponding to DFT size N = 12 into cluster #0 having partially orthogonal bandwidth B' (=N'*Bsub-9Bsub) corresponding to vector length N'=9 and cluster #1 having bandwidth B" (=N"*BSub=3BSub) corresponding to vector length N"=3. Here, in cluster #0 of vector length N'-9 which is multiple aoxo of prime number x0=3, there is a column vector which is orthogonal (hierarchically orthogonal) in vector length 3, 6, 9. For example, in real parts and imaginary parts of column vectors fo to f11 shown in FIG.6, their respective waveforms are orthogonal to each other in vector length 3, 6, 9 between column vectors fo and f4, between column vectors fo and fg, and between column vectors f4 and f8. Here, only an orthogonal relationship among vector lengths which are multiples of prime number xo=3 is shown. For example, between column vectors f4 and fg, vector length 3 matches a one-cycle portion of column vector f4 and a two-cycle portion of column vector fg, vector length 6 matches a two-cycle portion of column vector f4 and a four-cycle portion of column vector fg and vector length 9 matches a three-cycle portion of column vector f4 and a six-cycle portion of column vector fg, That is, column vectors fo, f4 and fg of 12 column vectors fo to fn in cluster #0 (vector length N'=9) have a hierarchically orthogonal relationship in which those column vectors are orthogonal to each other in a cycle of vector length 3, 6, 9. Thus, in cluster #0 (vector length N'=9), 1SI is reduced between column vectors fo, f4 and f8 (e.g. multiplexed symbols #0, #4, #8) of 12 column vectors f0 to f11 (e.g. multiplexed symbols #0 to #11) shown in FIG.6. Thus, according to the present division method, division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' which is multiple a0x0 of prime number Xo, and can thereby generate a cluster including more multiplexed symbols which are hierarchically orthogonal in a cycle of a multiple (x0, 2x0) ... , aoxo) of prime number Xo. That is, it is possible to produce more multiplexed symbols (column vectors) which are partially orthogonal to each other in cluster size of clusters generated by dividing the SC-FDMA signal. In other words, by reducing multiplexed symbols (column vectors) which are not partially orthogonal to each other in cluster size of clusters generated by dividing the SC-FDMA signal, it is possible to reduce ISI caused by the loss of orthogonality between multiplexed symbols which are not partially orthogonal to each other. Furthermore, according to the present division method, coefficient ao is the only information that needs to be reported from the base station to terminal 100 as control information on the division of the SC-FDMA signal (spectrum), and it is thereby possible to reduce the amount of information required to report the control information. A case has been described in the present division method where division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' which is a multiple of one prime number. However, in the present invention, for example, division section 111 may also divide the SC-FDMA signal with partially orthogonal bandwidth ET corresponding to vector length N' which is a multiple of a product of two or more prime numbers. For example, division section 111 assumes vector length N' to be a multiple (e.g. bo(xo*xi)) (where bo is an integer equal to or greater than 1) of a product (e.g. x0*x1) of at least two prime numbers (two or more prime numbers) of prime numbers xo, x1, x2, ... . Thus, the cluster having partially orthogonal bandwidth B' corresponding to vector length N'=bo(xo*X]) can include multiplexed symbols (column vectors) which are hierarchically partially orthogonal to each other in a cycle of a multiple (x0, 2x0, ... , b0x0) of prime number x0 and multiplexed symbols (column vectors) which are hierarchically partially orthogonal to each other in a cycle of a multiple (x1, 2x1; ... , box1) of prime number x1 That is, as the minimum division unit (e.g. xo*xi) of the SC-FDMA signal increases, it is possible to increase the number of multiplexed symbols (column vectors) which are partially orthogonal to each other in cluster size with the cluster having partially orthogonal bandwidth B' corresponding to vector length N' = bo(xo*X1). It is thereby possible to further reduce ISI caused by the loss of orthogonality between multiplexed symbols (column vectors). When two or more prime numbers are selected, it is preferable to select prime numbers in order from a smaller prime number (2, 3, 5, 7, ...). Thus, it is possible to produce more multiplexed symbols (column vectors) which are hierarchically orthogonal to each other in a cycle of a multiple of a prime number in a cluster having partially orthogonal bandwidth B' and further reduce ISI caused by the toss of orthogonality between multiplexed symbols (column vectors). In the present division method, division section 111 divides an SC-FDMA signal having partially orthogonal bandwidth B' corresponding to vector length N' which is a power of a prime number. This will be described more specifically below. For example, division section 111 assumes column vector length N! to be power xoa0 (where ao is an integer equal to or greater than I) of prime number x0. Here, suppose DFT size N is 12 and the number of clusters is 2 as in the case of division method 1-3. Furthermore, suppose prime number Xo=2 and coefficient ao=3. Thus, for example, division section 111 assumes vector length N' of cluster #0 to be 8 (=23) and assumes vector length N" of cluster #1 to be 4 (=N-N' = 12-8). That is, division section 111 divides an SC-FDMA signal having orthogonal bandwidth B (="N*Bsub= 12Bsub) corresponding to DFT size N=12 into cluster #0 having partially orthogonal bandwidth B' (=N'*Bsub=8Bsub) corresponding to vector length N'=8 and cluster #1 having bandwidth B" (-N"*Bsub=4Bsub) corresponding to vector length N"=4. Here, there are column vectors which are orthogonal to each other in vector lengths of 2, 4, 8 in cluster #0 having vector length N'=8 which is power x0a0 of prime number x0=2. For example, in real parts and imaginary parts of column vectors fo to fn shown in FIG.7, their respective waveforms are orthogonal to each other in vector length 2, 4, 8 between column vectors fo and f3, between column vectors fo and f6 and between column vectors f3 and f6 as in the case of division method 1-3 (FIG.6). Here, only an orthogonal relationship between vector lengths which are powers of prime number xo=2 is shown. That is, column vectors f0, f3, fe of 12 column vectors f0 to fn in cluster #0 (vector length N'=8) have a hierarchic orthogonal relationship in which those column vectors are orthogonal to each other in a cycle of vector length 2, 4, 8. Thus, in cluster #0 (vector length N'=8), ISI is reduced between column vectors f0, f3, f6 (e.g. multiplexed symbols #0, #3, #6) of 12 column vectors fo to f11 (e.g. multiplexed symbols #0 to #11) shown in FIG.7. Thus, according to the present division method, division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' which is power xoa0 of prime number xo, and can thereby generate clusters including more multiplexed symbols (column vectors) which are hierarchically orthogonal in a cycle of a power (xo, xo2, ... , xoa0) of prime number XO. Thus, it is possible to reduce ISI caused by the loss of orthogonality between multiplexed symbols (column vectors) which are not partially orthogonal to each other in cluster size of clusters generated by dividing the SC-FDMA signal as in the case of division method 1-3. Furthermore, according to the present division method, coefficient ao is the only information that needs to be reported from the base station to terminal 100 as control information on the division of the SC-FDMA signal (spectrum) and it is thereby possible to reduce the amount of information required to report the control information as in the case of division method 1-3. A case has been described in the present division method where division section 111 divides the SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' which is a power of one prime number. However, in the present invention, for example, division section 111 may also divide the SC-FDMA signal with a partially orthogonal bandwidth B: corresponding to vector length N' which is a power of a product of two or more prime numbers. For example, division section 111 assumes vector length N' to be a power (e.g. (x0*x1)b0) (where b0 is an integer equal to or greater than 1) of a product (e.g. x0*X1) of at least two prime numbers (two or more prime numbers) of prime numbers x0, x1, X2, ... . Thus, the cluster having partially orthogonal bandwidth B! corresponding to vector length N'=(xo*X1)b0 can include multiplexed.symbols (column vectors) which are hierarchically partially orthogonal to each other in a cycle of a power (x0, x02, ... , x0b0) of prime number x0 and multiplexed symbols (column vectors) which are hierarchically partially orthogonal to each other in a cycle of a power (xi; x12, ... , x1b0) of prime number x1. That is, as the minimum division unit (e.g. xo*xi) of the SC-FDMA signal increases, it is possible to increase the number of multiplexed symbols (column vectors) which are partially orthogonal to each other in cluster size of the cluster having partially orthogonal bandwidth B' corresponding to vector length N''=xo+x1)bo. It is thereby possible to further reduce ISI caused by the loss of orthogonality between multiplexed symbols (column vectors). Furthermore, in the present invention, division section 111 may also assume vector length N' to be a multiple (e.g. po(xo*x1)b0)) (where po is an integer equal to or greater than 1) of a power (e.g. (xo*xj) ) of a product (e.g. xo*X1) of at least two prime numbers (two or more prime numbers) of prime numbers xo, x1, X2, ... . Effects similar to those of the present division method may be obtained in this case, too. Furthermore, in the present invention, division section 111 may also assume vector length N' to be product xoc0*xicl* ... of at least two (two or more) powers x0e0, XiCl, ... (Co, C|, ... is an integer equal to or greater than 0, where, at least one of Co, C|, ... is an integer equal to or greater than 1) of prime numbers xo, x1, ... . Effects similar to those of the present division method may be obtained in this case, too. Here, in FFT (Fast Fourier Transform) that realizes processing equivalent to that of DFT by a smaller amount of calculations, a product of a power of a certain value may be used as the FFT size (the number of FFT points). Thus, when using FFT as a substitute for DFT, it is possible to improve affinity between FFT processing and division processing of the SC-FDMA signal by using a product of powers of prime numbers xoc0*X|Cl* ... as vector length N! for dividing column vector length N. Furthermore, division section 111 may also assume vector length N' to be multiple po (x0c0*X1Cl* ...) (where po is an integer equal to or greater than 1) of a product of powers of prime numbers xoc0*Xtcl* ... . When two or more prime numbers are selected, it is preferable to select prime numbers in order from a smaller prime number (2, 3, 5, 7, ...). It is thereby possible to produce more multiplexed symbols (column vectors) which are hierarchically partially orthogonal to each other in a cycle of a power of a prime number in clusters having partially orthogonal bandwidth B' and further reduce ISI caused by the loss of orthogonality between multiplexed symbols (column vectors). SC-FDMA signal division methods 1-1 to 1-4 through division section 111 have been described so far. Thus, even when dividing an SC-FDMA signal into a plurality of clusters and mapping the plurality of clusters to discontinuous frequency bands respectively, the present embodiment can reduce ISI caused by the loss of orthogonality of the DFT matrix by dividing the SC-FDMA signal with a partially orthogonal bandwidth. Thus, the present embodiment reduces I SI caused by the loss of orthogonality of the DFT matrix, and can thereby improve transmission characteristics without deteriorating data transmission efficiency even when using high-level M-ary modulation such as 64 QAM which has a very short Euclidian distance between signal points. A case has been described in the present embodiment where a terminal divides an SC-FDMA signal into a plurality of clusters so that a bandwidth of one cluster (here, cluster #0) is a partially orthogonal bandwidth. However, the terminal in the present invention may also divide the SC-FDMA signal into a plurality of clusters using one of division methods 1-1 to 1-4 so that bandwidths of all of the plurality of clusters are partially orthogonal bandwidths. Thus, it is possible to increase the number of multiplexed symbols having a partially orthogonal relationship with each other in all clusters and thereby reduce ISI cluster by cluster. Furthermore, in the present embodiment, the terminal may perform frequency interleaving for each frequency band (or cluster) having a partially orthogonal bandwidth as shown in FIG.8. To be more specific, when division section 111 divides the SC-FDMA signal into cluster #0 and cluster #1 as shown in the upper part of FIG.8, an interleaving section (not shown) performs frequency interleaving in units of partially orthogonal bandwidth. That is, the interleaving section performs frequency interleaving on a first-half portion of cluster #0 having partially orthogonal bandwidth Bo', a last-half portion of cluster #0 having partially orthogonal bandwidth B0' and cluster #1 having partially orthogonal bandwidth B1'. Thus, it is possible to further improve the frequency diversity effect while reducing the loss of orthogonality in the clusters as in the case of the present embodiment. Furthermore, a case has been described in the present embodiment where the base station reports only frequency resource information to terminal 100 every time the base station communicates with terminal 100 and terminal 100 calculates cluster information (the number of clusters and the cluster size) based on category information and partially orthogonal condition information (equation 1 and equation 2) reported beforehand. However, in the present invention, for example, the base station may report all frequency resource information and cluster information (the number of clusters and the cluster size) to terminal 100 every time the base station communicates with terminal 100 and terminal 100 may divide the SC-FDMA signal based on the received frequency resource information and cluster information. Furthermore, for example, the base station may also report frequency resource information showing frequency bands allocated in consideration of the number of clusters and the cluster size to terminal 100. To be more specific, the base station (scheduler of the base station) performs scheduling and thereby performs allocation processing of allocating frequency bands of partially orthogonal bandwidth B' that includes a frequency band of terminal 100 showing a maximum SINR in a certain frequency band (subcarrier) and satisfies equation 2 (or equation 1) on terminal 100. That is, the base station allocates frequency bands of partially orthogonal bandwidth B' calculated according to equation 2 (or equation 1) to a plurality of clusters constituting a C-SC-FDMA signal of terminal 100. The base station allocates frequency resources of the C-SC-FDMA signal made up of a plurality of clusters having a partially orthogonal bandwidth by repeatedly performing the above described allocation processing in different frequency bands. The base station then reports frequency resource information showing the frequency resource allocation result of the C-SC-FDMA signal of terminal 100 to terminal 100. The base station also performs the above described frequency resource allocation processing on terminals other than terminal 100. This allows the base station to schedule the allocation of frequency resources to all terminals locating in the cell of the base station. Furthermore, terminal 100 may map the C-SC-FDMA signal according to the frequency band shown in the frequency resource information reported from the base station. This allows terminal 100 to divide SC-FDMA into a plurality of clusters, map the plurality of clusters to frequency bands having a partially orthogonal bandwidth and can thereby have effects similar to those of the present embodiment. (Embodiment 2) The present embodiment will describe a case where MIMO (Multi-Input Multi-Output) transmission, which is one of transmission techniques for realizing high-speed, large-volume data transmission, is used. The MIMO transmission technique provides a plurality of antennas for both a base station and a terminal, provides a plurality of propagation paths (streams) in a space between radio transmission/reception, spatially multiplexes the respective streams, and can thereby increase throughput. This will be described more specifically below. FIG.9 shows a configuration of terminal 200 according to the present embodiment. Terminal 200 is provided with two antennas (antennas 101-1 and 101-2) that transmit C-SC-FDMA signals (a plurality of clusters) using two streams (stream #1 and stream #2). Furthermore, terminal 200 includes C-SC-FDMA processing sections 201-1 and 201-2 made up of coding section 107, modulation section 108, multiplexing section 109, DFT section 110 and division section 111, respectively provided for antennas 101-1 and 101-2. Furthermore, terminal 200 also includes transmission processing sections 203-1 and 203-2 made up of mapping section 112, 1FFT section 113, CP insertion section 1)4 and radio transmitting section 115, respectively provided for antennas 101-1 and 101-2. C-SC-FDMA processing sections 201-1 and 201-2 generate C-SC-FDMA signals (a plurality of clusters) by applying processing similar to that by coding section 107 to division section 111 in Embodiment 1 to transmission bit sequences inputted respectively. C-SC-FDMA processing sections 201-1 and 201-2 then output the C-SC-FDMA signals generated to precoding section 202 respectively. Precoding section 202 receives different spatial precoding matrixes (PM) for each identical frequency band having a partially orthogonal bandwidth or for each identical cluster of the partially orthogonal bandwidth from control section 106 as input. That is, precoding section 202 uses the same spatial precoding matrix for each identical frequency band having a partially orthogonal bandwidth or for each identical cluster having a partially orthogonal bandwidth. Here, precoding information showing the spatial precoding matrix is reported from a base station to terminal 200. For example, the precoding information shows a number indicating each spatial precoding matrix and control section 106 may calculate each spatial precoding matrix based on the number indicated in the precoding information. Precoding section 202 multiplies the C-SC-FDMA signals inputted from C-SC-FDMA processing sections 201-1 and 201-2 by the spatial precoding matrix respectively. Here, precoding section 202 multiplies the C-SC-FDMA signals mapped to frequency bands having the same partially orthogonal bandwidth or clusters having the same partially orthogonal bandwidth by the same spatial precoding matrix in each of the plurality of streams. Precoding section 202 then outputs the precoded C-SC-FDMA signals to corresponding transmission processing sections 203-1 and 203-2 for each stream. Transmission processing sections 203-1 and 203-2 apply processing similar to that of mapping section 112 to radio transmitting section 115 of Embodiment 1 to the precoded C-SC-FDMA signals inputted respectively and transmit the C-SC-FDMA signals after the transmission processing to the base station via antennas 101-1 and 101-2 respectively. Next, details of the precoding processing by precoding section 202 of terminal 200 will be described, First, a case will be described where the same spatial precoding matrix is used for each partially orthogonal band. For example, in FIG.I0A, each division section 111 (FIG.9) of C-SC-FDMA processing sections 201-1 and 201-2 divides an SC-FDMA signal into cluster #0 having a bandwidth twice partially orthogonal bandwidth B0' and cluster #1 having partially orthogonal bandwidth B1 Therefore, precoding section 202 multiplies cluster #0 and cluster #1 transmitted by the same spatial precoding matrix for every partially orthogonal bandwidth using stream #1 and stream #2. To be more specific, as shown in FIG.10A, precoding section 202 uses the same spatial precoding matrix PM #0 for both stream #1 and stream #2 in one partially orthogonal bandwidth B0' of cluster #0 and uses the same spatial precoding matrix PM #1 for both stream #1 and stream #2 in the other partially orthogonal bandwidth Bo'. Furthermore, precoding section 202 uses the same spatial precoding matrix PM #2 for both stream #1 and stream #2 in cluster #1 having partially orthogonal bandwidth Bi'. Next, a case will be described where the same spatial precoding matrix is used for each cluster. For example, in FIG.1 OB, each division section 111 (FIG.9) of C-SC-FDMA processing sections 201-1 and 201-2 divides an SC-FDMA signal into cluster #0 having partially orthogonal bandwidth Bo' and cluster #1 having partially orthogonal bandwidth Bj'. Precoding section 202 then multiplies cluster #0 and cluster #1 transmitted using stream #1 and stream #2 by the same spatial precoding matrix for each cluster. To be more specific, as shown in FIG. 10B, precoding section 202 uses the same spatial precoding matrix PM #0 for both stream #1 and stream #2 in cluster #0 having partially orthogonal bandwidth Bo'. Furthermore, precoding section 202 uses the same spatial precoding matrix PM #2 for both stream #1 and stream #2 in cluster #1 having partially orthogonal bandwidth Bi'. Thus, for example, in FIG.10A, between cluster #0 of stream #1 and cluster #1 of stream #2, it is possible to reduce 1SI by maintain orthogonality between multiplexed symbols (column vectors) in the respective clusters in the frequency domain as in the case of Embodiment 1, while in the spatial domain, it is possible to maintain orthogonality between them using spatial precoding matrixes (e.g. unitary matrixes) orthogonal to each other. That is, it is possible to further reduce IS1 between cluster #0 of stream #1 and cluster #1 of stream #2 (that is, between clusters transmitted with different frequency bands and different streams). The same applies between cluster #1 of stream #1 and cluster #0 of stream #2. That is, when using the MIMO transmission technique, it is possible to reduce ISI between different streams and between different frequency bands by using the same spatial precoding matrix for each identical partially orthogonal bandwidth (or each cluster) in different streams. By this means, the present embodiment can reduce ISI in the frequency domain by dividing the SC-FDMA signal with a partially orthogonal bandwidth as in the case of Embodiment I and further reduce ISI in the spatial domain by using a spatial precoding matrix for each partially orthogonal bandwidth. Although a case has been described in the present embodiment where two streams are used, the number of streams is not limited to two but the present invention may also be applied to cases where three or more streams are used. Furthermore, the present embodiment is applicable to both single user (SU)-MIMO transmission (that is, MIMO transmission between a plurality of antennas of one base station and a plurality of antennas of one terminal) and multiuser (MU)-MIMO transmission (that is, MIMO transmission between a plurality of antennas of one base station and a plurality of antennas of a plurality of terminals). Furthermore, in the present embodiment, when FSTD (Frequency Switched Transmit Diversity) is used, the terminal may switch between transmitting antennas for each frequency band (or cluster) having a partially orthogonal bandwidth. For example, as shown in FIG.11, when the number of transmitting antenna is 3 (antennas #0 to #2) and the number of clusters is 3 (clusters #0 to #2), the first half part of cluster #0 having partially orthogonal bandwidth B0' may be transmitted from antenna #0, the second half part of cluster #0 having partially orthogonal bandwidth B0' may be transmitted from antenna #1, cluster #1 having partially orthogonal bandwidth B|! may be transmitted from antenna #0 and cluster #2 having partially orthogonal bandwidth B2' may be transmitted from antenna #2. Thus, by switching between transmitting antennas based on the unit of frequency bands (or clusters) having a partially orthogonal bandwidth in FSTD, it is possible to receive a fading variation which differs among frequency bands (Bo' to B2') having partially orthogonal bandwidths. Therefore, it is possible to obtain a space diversity effect while maintaining orthogonality within a frequency band having partially orthogonal bandwidths. (Embodiment 3) A case has been described in Embodiment 2 where when FSTD (Frequency Switched Transmit Diversity) is used, a terminal switches between transmitting antennas for each frequency band (or cluster) having a partially orthogonal bandwidth. Furthermore, in this case, a case has been described where a plurality of clusters are mapped to non-continuous frequency bands when viewed in the frequency domain of all transmitting antennas. By contrast, in the present embodiment, when using FSTD that switches between transmitting antennas for each frequency band (or cluster) having a partially orthogonal bandwidth, a terminal maps a plurality of clusters to continuous frequency bands when viewed in the frequency domain of all transmitting antennas. That is, when FSTD is used in Embodiment 2, as shown in F1G.11, clusters having partially orthogonal bandwidths mapped to the respective antennas are mapped to non-continuous frequency bands and a plurality of clusters are mapped to non-continuous frequency bands when also viewed in frequencies of all antennas. To be more specific, there is an inter-antenna vacant frequency band between cluster #0 of antenna #1 and cluster #1 of antenna #0 in FIG.11. Likewise, there is also an inter-antenna vacant frequency band between cluster #1 of antenna #0 and cluster #2 of antenna #2. Furthermore, in FIG.11, no cluster is mapped to any inter-antenna vacant frequency band and a plurality of clusters are mapped to non-continuous frequency bands when also viewed in the frequency domain of all antennas. On the other hand, in the present embodiment, when FSTD is used, as shown in FIG.12, clusters having partially orthogonal bandwidths to be mapped to the respective antenna (space resources) are mapped to non-continuous frequency bands as in the case of Embodiment 2. On the other hand, as shown in FIG.12, a plurality of clusters having partially orthogonal bandwidths to be mapped to the respective antennas (space resources) are mapped to continuous frequency bands when viewed in the frequency domain of all antennas. That is, in FIG.12, there is no vacant frequency band between any clusters; between cluster #A of antenna #0 (space resource #0) and cluster #B of antenna #1 (space resource #1), between cluster #B of antenna #1 (space resource #1) and cluster #C of antenna #0 (space resource #0) and between cluster #C of antenna #0 (space resource #0) and cluster #D of antenna #2 (space resource #2). That is, when viewed in the frequency domain of all antennas, a plurality of clusters having partially orthogonal bandwidths are mapped to continuous frequency bands. That is, when viewed in the frequency domain of each antenna, even when C-SC-FDMA signals (a plurality §f clusters having partially orthogonal bandwidths) are mapped to non-continuous frequency bands, if C-SC-FDMA signals are mapped to continuous frequency bands when viewed in the frequency domain of all antennas, it is possible to further obtain space diversity effects while maintaining orthogonality within a frequency band having partially orthogonal bandwidths as in the case of Embodiment 2. Furthermore, the receiving apparatus (base station) side can perform reception processing in the same way as when the transmitting apparatus (terminal) side transmits SC-FDMA signals to continuous frequency bands. Thus, according to the present embodiment, the receiving apparatus (base station) can obtain space diversity effects while maintaining orthogonality within a frequency band of partially orthogonal bandwidths without being aware of non-continuous mapping processing between antennas (between space resources) of the transmitting apparatuses. The present invention may also use a method of mapping a plurality of clusters having partially orthogonal bandwidths so as to rotate the antenna axis (or antenna direction, space resource region) in the frequency domain as the method of mapping the plurality of clusters having partially orthogonal bandwidths to the plurality of antennas. FIG.13 shows a case where the terminal maps a plurality of clusters (clusters #A, #B, #C, #D) to antennas #0 to #2 (space resources #0 to #2) in such a way that the clusters rotate in the same direction of the antenna axis (or antenna direction, space resource region) in order from a low frequency to a high frequency. To be more specific, as shown in FIG.13, the terminal maps cluster #A to antenna #0 (space resource #0), maps cluster #B to antenna #1 (space resource #1), maps cluster #C to antenna #2 (space resource #2) and maps cluster #D to antenna #0 (space resource #0). That is, in FIG.13, the terminal maps clusters #A, #B, #C and #D so as to rotate in the same direction of the antenna axis (or antenna direction. space resource region) (that is, in the rotating direction in which the antenna number (space resource number) cyclically increases as the frequency increases) in order of antennas #0, #1, #2, #0, ... . Furthermore, as shown in FIG.13, four clusters #A, #B, #C and #D are mapped to continuous frequency bands when viewed in the frequency domain of all antennas as in the case of FIG. 12. Thus, since the frequency domain of antennas (space resources) to which a plurality of clusters are mapped is set cyclically, only one piece of frequency resource allocation information (continuous frequency resources or non-continuous frequency resources) needs to be reported to the plurality of antennas as frequency resource allocation information when the plurality of clusters are mapped to the frequency domain of the plurality of antennas. Thus, it is possible to obtain effects similar to the present embodiment while reducing the amount of information required to allocate frequency resources to the respective antennas. By sharing information on the rotating direction on the antenna axis (space resource region) (e.g. the rotating direction in which the antenna number (space resource number, layer number) cyclically increases (decreases) as the frequency increases (decreases)) between the base station and the terminal, only one piece of frequency resource allocation information needs to be reported to the plurality of antennas as control information from the base station to the terminal. FIG.13 has described a case with the rotating direction in which the antenna number (space resource number) of the antenna to which each cluster is mapped cyclically increases as the frequency increases as an example. However, in the present invention, the rotating direction of the antenna axis (space resource region) in the frequency domain may also be a rotating direction in which the antenna number (space resource number, layer number) cyclically decreases as the frequency increases. Furthermore, the rotating direction of the antenna axis (space resource region) may also be switched for every certain frequency band (subband unit made up of a plurality of subcarriers, resource block unit or resource block group unit or the like). Alternatively, the rotating direction of the antenna axis (space resource region) may also be switched for every certain time unit (symbol unit, slot unit, subframe unit or number of retransmissions is performed or the like). Alternatively, the rotating direction of the antenna axis (space resource region) may also be switched for every certain time-frequency unit made up of two-dimensional resources of the time domain and the frequency domain. For example, a frequency band allocated to a terminal may be divided into two portions and a plurality of clusters having partially orthogonal bandwidths may be mapped to a plurality of antennas in the rotating direction in which the antenna number of an antenna to which each cluster is mapped cyclically increases as the frequency increases in one frequency band and in the rotating direction in which the antenna number of an antenna to which each cluster is mapped cyclically decreases as the frequency increases in the other frequency band. Furthermore, when, for example, one codeword made up of a plurality of symbols is mapped over two slots (e.g. first slot and second slot), a plurality of clusters having partially orthogonal bandwidths may be mapped to a plurality of antennas in the rotating direction in which the antenna number of an antenna to which each cluster is mapped cyclically increases as the frequency increases in the first slot and in the rotating direction in which the antenna number of an antenna to which each cluster is mapped cyclically decreases as the frequency increases in the second slot. Thus, it is possible to increase randomness of channels in the frequency domain (or time domain) while maintaining a partially orthogonal relationship in each cluster and thereby further improve the diversity effect. Furthermore, a case has been described in FIG. 13 where the antenna number of an antenna to which each cluster is mapped is rotated in the same direction of the antenna axis (or antenna direction, space resource region) in order from a lowest frequency and a plurality of clusters are mapped to the antennas (space resources). However, the present invention may also be adapted so that the antenna number of an antenna to which each cluster is mapped is rotated in the same direction of the antenna axis (or antenna direction, space resource region) in order from a higher frequency and a plurality of clusters are mapped to the antennas (space resources). Furthermore, a case has been described in FIG. 13 where the terminal maps the clusters to a plurality of antennas over continuous frequency bands while rotating four clusters #A to #D among different antennas (antennas #0 to #2) as an example. However, in the present invention, the terminal may also map the clusters to non-continuous frequency bands over a plurality of antennas while rotating the plurality of clusters among different antennas in the same way as in FIG. 11. That is, in FIG.13, there may be a vacant frequency band (frequency band to which no cluster is allocated) between any clusters; between cluster #A of antenna #0 and cluster #B of antenna #1, between cluster #B of antenna #1 and cluster #C of antenna #2 and between cluster #C of antenna #2 and cluster #D of antenna #0. (Embodiment 4) of Embodiment 1 has described a case where division section 111 (FIG.I) divides an SC-FDMA signal with partially orthogonal bandwidth B' corresponding to vector length N' in (1) to (5) shown below. (1) Power of prime number x0: N'=xoa0 (where ao is an integer equal to or greater than 1) (2) Power of a product of at least two prime numbers (two or more prime numbers) of prime numbers x0, x1, x2, ... : N'=(xo*Xi)bo (where bo is an integer equal to or greater than 1) (3) A multiple of a power of a product of at least two prime numbers (two or more prime numbers) of prime numbers XO XI, x2, ... : N'=po(Xo*X1)b0 (where po is an integer equal to or greater than 1) (4) A product of at least two (two or more) of powers x0c0, xtcl, ... (co, C1, ... is an integer equal to or greater than 0, however at least one of c0, C[, ... is an integer equal to or greater than I) of prime numbers xo, xi, ... : N'=x0c0*x,cl* ... (5) A multiple of a product of powers of prime numbers xoc0*X1Cl* ... : N'=po(xoc0*X1Cl* ...) (where po is an integer equal to or greater than 1) Here, a product of prime numbers (e.g. (xo*xi)) or a product of powers of prime numbers (e.g. (xoc0*x1cl)) is represented by a finite number of values equal to or greater than 2 (e.g. two numerical values of xo and X1 or two numerical values of x0c0 and X1C|). That is, when a prime number which is the base of a power is represented by Xi (i=0 to M-l) and the exponent of the power is represented by Cj (i=0 to M-l), M becomes a finite value showing an integer of 2 or more. The present embodiment is different from in Embodiment 1 in that coefficients of powers (that is, exponents of powers) Co, C1, ... , CM-I are made related to the bases of the powers (that is, prime numbers) x0, X1, ... , XM-I in the division method using vector length N; in above (4) and vector length N' in (5) described in of Embodiment I. To be more specific, when the base (prime number) of power is represented by X1 (i=0 to M-l) and the exponent of the power thereof is represented by C1 (i=0 to M-l), control section 106 (FIG.I) of terminal 100 according to the present embodiment sets the value of C1 corresponding to X1 to a value equal to or smaller than the exponent of the power having a greater base for the product of powers xoc0*X1Cl* ... *xM-icM"1 as the value of Xj increases. That is, when the base (prime number) of the power has a relationship of XiCi (i≠i). Therefore, when the bases of power have a relationship of xoC1>c2> ... >CM-I- Control section 106 calculates vector length N'=xoc0*X1Cl* ... *xM.|CM-1 (corresponds to vector length N' in (4) of ) or vector length N'=p0(xoc0*x1c1* ... *xM-1cM"') (corresponds to vector length N' in (5) of ). Division section 111 then divides the SC-FDMA signal with vector length N' or partially orthogonal bandwidth B' corresponding thereto. That is, division section III divides the SC-FDMA signal with a partially orthogonal bandwidth corresponding to vector length N' where value of exponent c, of certain power xici (i is one of 0 to (M-l)) among a plurality of powers (x0c0, xicl, ... , xM.1cM"') constituting a product (x0c0*X1c1* ... *XM-ICM"1) of powers representing vector length N' becomes equal to or smaller than value of exponent c;> of another power Xi'C1 having a smaller base than base Xj of the certain power XiCl (that is, a power corresponding to Xi'Xi, where i≠i). Mapping section 112 maps the plurality of clusters generated by dividing the SC-FDMA signal to non-continuous frequency bands. Thus, it is possible to increase the number of combinations of partially orthogonal column vectors having a shorter cycle in each cluster of a partially orthogonal band (length) represented by equation 1 and equation 2 and thereby further reduce I SI. Hereinafter, a case will be described as an example where vector length N' (=xoc0*x,cl* ... *x,cM-') in (4) of division method l-4> of Embodiment 1 is used. Here, suppose M=3 and the base of each power is xo=2, X]=3, x2=5 (that is, xoCi>c2 (example 2, that is, the present embodiment). First, a case with c0=0, C| = 1, c2=2 (co in Embodiment 1, the present embodiment can further reduce ISI caused by the loss of orthogonality of the DFT matrix in the cluster. In the present invention, the division method using the relationship between the base of the power (x0C1>c2> ... >cM-1) may be applied to all cluster sizes. When, for example, two clusters are generated from an SC-FDMA signal (spectrum) generated through DFT processing with N=420 points, the terminal may divide the SC-FDMA signal after setting the cluster sizes of the two clusters to 360 and 60 respectively and map the two clusters to non-continuous bands. Here, since 360 and 60 can be expressed by 360=23*32*5' and 60=22*315', both cluster sizes satisfy the condition (relationship between the base of the power (xo:C1>C2> ... >CM-I)) in the present embodiment. This makes it possible to increase the number of column vectors of the DFT matrix having a partially orthogonal relationship in all clusters and thereby further reduce ISI caused by the loss of orthogonality of the DFT matrix in all non-continuously allocated bands. Furthermore, in the present invention, when, for example, the base of the power becomes x0ci> ... >cM-1, the terminal can set vector length N' (=xoc0*X1Cl* ... *XM--ICM _12) greater than minimum division unit X, it is possible to create a number of partially orthogonal relationships greater than the number of column vectors which have a partially orthogonal relationship in the length of minimum division unit X between column vectors in the cluster. That is, it is possible to secure an 1S1 reduction effect obtained by minimum division unit X in all clusters generated by dividing the SC-FDMA signal. Furthermore, by sharing minimum division unit X between the base station and the terminal in this case, only multiplier po may be reported from the base station to the terminal (or from the terminal to the base station) as control information on the division. This allows the amount of information required to report the control information to be reduced. Furthermore, when setting minimum division unit X (vector length N')=x0c0*x1cl* ... *xM-1cM'''(Xj, where i ≠ i) This allows a relationship of to be created with a cluster whose length (bandwidth) can be represented by That is, in a cluster having a length (bandwidth) of poX, it is possible to increase the number of combinations of column vectors which have a shorter cycle and are hierarchically partially orthogonal to each other. This makes it possible to create partially orthogonal relationships between column vectors of the DFT matrix even in a cycle of a power of Xi (i=0 to M'-l) in all clusters generated by dividing an SC-FDMA signal and thereby further reduce ISI. FIG.14 shows cluster size N! assuming M=3 and minimum division unit (that is, wherein multiplier las a relationship of and(where M'=3), FIG.14 shows a case with M=M' (=3) as an example, but M^M' may also be applicable. For example, in the case with number #3 shown in FIG. 14, since multiplier cluster size satisfying a relationship of That is, in a cluster of vector length N' = 72, it is possible to create combinations of column vectors which have a shorter cycle such as 2, 3, 4, 6, 8, 9, ... and in which column vectors of the DFT matrix are made to be hierarchically partially orthogonal in lengths of a power of 2, power of 3, power of 4, ... . Furthermore, as described in of Embodiment 1, when the SC-FDMA signal is divided with partially orthogonal bandwidth B' corresponding to vector length N' which is a multiple of a prime number (N'-aoXo (where the prime number is x0, coefficient ao is an integer equal to or greater than 1)), that is, when the SC-FDMA signal is divided assuming that x0 is a minimum division unit and that the cluster size of each cluster is a length corresponding to a multiple of the minimum division unit, the multiplier (coefficient ao) may be power x0d0 of prime number x0 (here, do is an integer equal to or greater than 0). This makes it possible to increase the number of combinations of column vectors which are hierarchically partially orthogonal in a cycle of a power of xo in a cluster having a length of aoXo(=x0d0+l) and thereby further reduce 1SI more than of Embodiment 1. Furthermore, as described in of Embodiment 1, when the SC-FDMA signal is divided with partially orthogonal bandwidth B' corresponding to vector length N' which is a multiple of a product of two or more prime numbers (e.g. N' = bo(xo*X1) (where x0 and X1 are prime numbers, coefficient bo is an integer equal to or greater than 1), that is, when the SC-FDMA signal is divided using (xo*X1) as a minimum division unit and assuming the size of each cluster to be a length corresponding to a multiple of the minimum division unit, the multiplier (coefficient bo) may be power (x0*X1)d0 of a product (xo*X|) of the prime numbers (here, d0 is an integer equal to or greater than 0). This makes it possible to increase the number of combinations of column vectors which are hierarchically partially orthogonal in a cycle of powers of xo, x1 and (xo*x1) of a cluster having a length of bo(x0*x1)(=(x0*x1)d0+l) and thereby further reduce ISI more than of Embodiment 1. (Embodiment 5) A case has been described in Embodiment 1 and Embodiment 4 where as shown in FIG. 1, the division section is connected to the DFT section of the terminal, the output signal (DFT output) of the DFT section is directly divided using the aforementioned division method and a plurality of clusters are thereby generated. By contrast, the present embodiment will describe a case where a shifting section is provided between the DFT section and the division section. To be more specific, the terminal according to the present embodiment causes the shifting section to cyclically frequency-shift DFT output (SC-FDMA signal (spectrum)) outputted from the DFT section, divide the SC-FDMA signal after the cyclical frequency shift among partially orthogonal bandwidths (lengths) and generate a plurality of clusters. FIG. 15 shows a configuration of a transmitting apparatus (terminal) according to the present embodiment. In terminal 300 shown in FIG.15, the same components as those in Embodiment 1 (FIG.1) will be assigned the same reference numerals and descriptions thereof will be omitted. Shifting section 301 receives a frequency domain signal (SC-FDMA signal) generated by applying DFT processing to a time domain symbol sequence from DFT section 110 as input and receives an amount of shift (amount of cyclic frequency shift) in a frequency domain set by the base station (or terminal 300) from control section 106 as input. Shifting section 301 then cyclically frequency-shifts the SC-FDMA signal inputted from DFT section 110 within a DFT band (DFT size N) in DFT processing by DFT section 110 according to the amount of cyclic frequency shift inputted from control section 106. That is, shifting section 301 applies cyclic frequency shift to the SC-FDMA signal within the DFT band. Shifting section 301 may also be configured so as not to cyclically frequency-shift the SC-FDMA signal (spectrum) of the pilot symbol of the sequence in which the data symbol and pilot symbol inputted to shifting section 301 are time-multiplexed. Shifting section 301 outputs the cyclically frequency-shifted SC-FDMA signal to division section 111. Details of the cyclic frequency shifting processing on the SC-FDMA signal (spectrum) by shifting section 301 will be described later. Division section 111 divides the cyclically frequency-shifted SC-FDMA signal inputted from shifting section 301 with partially orthogonal length (vector length) N' and generates a plurality of clusters using one of the division methods described in the aforementioned embodiments (e.g. Embodiment 1 or Embodiment 4), Next, FIG.16 shows the configuration of a receiving apparatus (base station) according to the present embodiment. Base station 400 shown in FIG.16 determines allocation of uplink frequency resources, parameters (cluster size and number of clusters or the like) about spectral division at each terminal and amount of cyclic frequency shift and reports the determined information to each terminal as information to be reported. Base station 400 may also report information on frequency resource allocation taking account of influences of spectral division and the amount of cyclic frequency shift based on parameters about spectral division to the terminal. Each terminal (terminal 300) then divides the cyclically frequency-shifted SC-FDMA signal (spectrum) based on parameters about spectral division included in the information reported from base station 400. In the configuration of receiving apparatus (base station 400) shown in FIG.16, the configuration except reverse shifting section 408, that is, the configuration in which an output signal from combining section 407 is directly inputted to 1DFT section 409 corresponds to the configuration of the receiving apparatus (base station) (not shown) of Embodiment 1. The receiving apparatus (base station 400) shown in FIG. 16 is comprised of antenna 401, radio receiving section 402, CP removing section 403, FFT section 404, demapping section 405, FDE section 406, combining section 407, reverse shifting section 408, IDFT section 409, demodulation section 410, decoding section 411, measuring section 412, scheduler 413, control section 414, generation section 415, coding section 416, modulation section 417 and radio transmitting section 418. Radio receiving section 402 of base station 400 receives an uplink C-SC-FDMA signal transmitted from each terminal via antenna 401 and applies reception processing such as down-conversion, A/D conversion to the C-SC-FDMA signal. Radio receiving section 402 outputs the C-SC-FDMA signal subjected to the reception processing to CP removing section 403. CP removing section 403 removes a CP added at the head of the C-SC-FDMA signal inputted from radio receiving section 402 and outputs the C-SC-FDMA signal after the removal of the CP to FFT (Fast Fourier Transform) section 404. FFT section 404 applies FFT to the C-SC-FDMA signal after the removal of the CP inputted from CP removing section 403 to convert the C-SC-FDMA signal to frequency domain C-SC-FDMA signals, that is, subcarrier components (orthogonal frequency components). FFT section 404 outputs the subcarrier components after the FFT to demapping section 405, Furthermore, when a subcarrier component after the FFT is a pilot signal, FFT section 404 outputs the subcarrier component to measuring section 412. Demapping section 405 demaps (extracts) a C-SC-FDMA signal (data signal) allocated to each subcarrier component (orthogonal frequency component) of a frequency resource used by a target terminal from the subcarrier components inputted from FFT section 404 based on frequency resource mapping information of the terminal inputted from control section 414. Demapping section 405 then outputs the demapped C-SC-FDMA signal to FDE section 406. FDE section 406 calculates an FDE weight based on an estimate value of a frequency channel gain between each terminal and base station 400 estimated by an estimation section (not shown) and equalizes the C-SC-FDMA signals inputted from demapping section 405 in the frequency domain using the calculated FDE weight. FDE section 406 then outputs the signal after the FDE to combining section 407. Combining section 407 combines the C-SC-FDMA signals (that is, C-SC-FDMA signals (spectra) after the FDE made up of a plurality of clusters) inputted from FDE section 406 in the frequency domain based on the cluster size and the number of clusters inputted from control section 414. Combining section 407 then outputs the combined C-SC-FDMA signal to reverse shifting section 408. Reverse shifting section 408 cyclically frequency-shifts in the direction opposite to the direction of shifting section 301 of terminal 300 (that is, reverse cyclic frequency-shifts) the combined C-SC-FDMA signal (spectrum) after the FDE according to the amount of cyclic frequency shift inputted from control section 414 (the same amount of cyclic frequency shift as the amount of cyclic frequency shift used by shifting section 301 of terminal 300). When, for example, the amount of cyclic frequency shift of shifting section 301 of terminal 300 is +z(-z), reverse shifting section 408 of base station 400 performs a -z(+z) cyclic frequency shift on-the combined signal after the FDE. Reverse shifting section 408 then outputs the C-SC-FDMA signal after the reverse cyclic frequency shift to IDFT section 409. IDFT section 409 applies IDFT processing to the C-SC-FDMA signal inputted from reverse shifting section 408 (C-SC-FDMA signal (spectrum) combined after the FDE and subjected to a reverse cyclic frequency shift) and thereby transforms the C-SC-FDMA signal to a time domain signal. IDFT section 409 then outputs the time domain signal to demodulation section 410. Demodulation section 410 demodulates the time domain signal inputted from IDFT section 409 based on MCS information (modulation scheme) inputted from scheduler 413 and outputs the demodulated signal to decoding section 411. Decoding section 411 decodes the signal inputted from demodulation section 410 based on MCS information (coding rate) inputted from scheduler 413 and outputs the decoded signal as a received bit sequence. On the other hand, measuring section 412 measures channel quality of each terminal in the frequency domain, for example, SINR (Signal-to-Interference plus Noise power Ratio) for each subcarrier of each terminal using pilot signals (pilot signals transmitted from each terminal) included in subcarrier components inputted from FFT section 404 and thereby generates channel quality information (CQI) of each terminal. Measuring section 412 then outputs the CQI of each terminal to scheduler 413. Scheduler 413 calculates priority of allocation of uplink shared frequency resources (PUSCH: Physical Uplink Shared CHannel) to each terminal using inputted information on QoS (Quality of Service) or the like of each terminal. Scheduler 413 then allocates each subcarrier (or frequency resource block RB (Resource Block) made up of a plurality of subcarriers) to each terminal using the calculated priority and the CQI inputted from measuring section 412. PF (Proportional Fairness) or the like may be used as an algorithm used to allocate frequency resources. Furthermore, scheduler 413 outputs frequency resource allocation information of each terminal showing frequency resources of each terminal allocated using the above described method to control section 414 and generation section 415 and outputs control information (MCS information or the like) other than the frequency resource allocation information to demodulation section 410, decoding section 411 and generation section 415. Control section 414 calculates the number of clusters and the cluster size of the terminal using the frequency resource allocation information of each terminal inputted from scheduler 413, category information of the terminal (information including the DFT size) and partially orthogonal condition information (information showing partially orthogonal condition (equation 1 or 2) of C-SC-FDMA). Furthermore, control section 414 calculates frequency resources to which C-SC-FDMA signals of each terminal are mapped based on the calculated number of clusters and cluster size. Control section 414 then outputs the calculated number of clusters and cluster size to combining section 407 and outputs the frequency resource mapping information showing frequency resources to which the C-SC-FDMA signals of each terminal are mapped to demapping section 405. Furthermore, control section 414 sets an amount of cyclic frequency shift used in reverse shifting section 408 and shifting section 301 of terminal 300 and outputs information on the set amount of cyclic frequency shift to reverse shifting section 408 and generation section 415. Generation section 415 converts the frequency resource allocation information inputted from scheduler 413, control information (MCS information or the like) other than the frequency resource allocation information and information on the amount of cyclic frequency shift inputted from control section 414 to a binary control bit sequence to be reported to each terminal and thereby generates a control signal. Generation section 415 then outputs the generated control signal to coding section 416. Coding section 416 codes the control signal inputted from generation section 415 and outputs the coded control signal to modulation section 417. Modulation section 417 modulates the control signal inputted from coding section 416 and outputs the modulated control signal to radio transmitting section 418. Radio transmitting section 418 applies transmission processing such as D/A conversion, amplification and up-conversion to the control signal inputted from modulation section 417 and transmits the signal subjected to the transmission processing to each terminal via antenna 401. Next, details of cyclic frequency shifting processing on an SC-FDMA signal (spectrum) by shifting section 301 of terminal 300 will be described. Since C-SC-FDMA performs precoding using a DFT matrix, even if DFT output (output signal of DFT processing) is cyclically shifted within a DFT band (DFT size N), it is possible to create a partially orthogonal relationship among column vectors at an arbitrary position of the DFT output as long as the cluster size of clusters generated through division is length N' that satisfies equation 1. The present embodiment takes advantage of this feature. This will be described more specifically below. That is, a feature in a section where column vectors of the DFT matrix are partially orthogonal to each other will be described. First, partially orthogonal conditions among column vectors of the DFT matrix in a segment of k=0 to N'-l of vector length N (section: k=0 to N-l) will be described. Two column vectors fj(k)(=fi) and fi' (k)(=fi-) (where i'≠i) having different angular frequencies in the DFT matrix are defined as following equation 3. and i, i'=0 to N-l. Here, of vector length N (section: k=0 to N-l), an inner product (partial cross correlation without time difference) of fj(k) and fj' (k) in partial vector length N' (segment: k=0 to N'-l) is as shown in following equation 4 (where N'