DESCRIPTION WIRELESS COMMUNICATION SYSTEM, WIRELESS COMMUNICATION APPARATUS, AND WIRELESS COMMUNICATION METHOD Technical Field In one aspect, the invention relates to a wireless communication system, a wireless communication apparatus, and a wireless communication method using spatial multiplexing. More particularly, the invention relates to a wireless communication system, a wireless communication apparatus, and a wireless communication method, in which a transmitter and a receiver share channel information to perform closed loop type spatial multiplexing transmission. In another aspect, the invention relates to a wireless communication system, a wireless communication apparatus, and a wireless communication method, which perform beamforming by obtaining a channel matrix on the basis of a training sequence transmitted from a receiver when a transmitter transmits a packet. More particularly, the invention relates to a wireless communication system, a wireless communication apparatus, and a wireless communication method, which perform beamforming using the training sequence transmitted from the transmitter to the receiver when the number of antennas of the transmitter, which is a beamformer, is smaller than that of the receiver, which is a beamformee. Background Art Wireless network has attracted much attention recently, because it is capable of removing wire in an existing wired communication network. Standard wireless networks include IEEE (Institute of Electrical and Electronics Engineers) 802.11 or IEEE 802.15. For example, IEEE 802 . lla/g, a standard of wireless Local Area Network (LAN), specifies an orthogonal frequency division multiplexing (OFDM) modulation method, which is a multi-carrier method. Because, in the OFDM modulation method, transmission data having orthogonal frequencies is distributed to a plurality of carriers and transmitted, the band of each carrier becomes narrow, spectrum efficiency is very high, and resistance to frequency-selective fading interference is strong. In addition, IEEE 802.lla/g standard supports a modulation method for accomplishing a communication speed up to 54 Mbps. However, a next-generation wireless LAN standard requires a higher bit rate. In order to realize a higher speed for wireless communications, multi-input multi-output (MIMO) communication has attracted attention. MIMO communication employs a plurality of antennas in a transmitter and in a receiver to realize spatially multiplexed streams. The transmitter performs spatial/temporal encoding and multiplexing of plural pieces of transmission data, and distributes and transmits the plural pieces of transmission data to N transmission antennas through channels, where N is a positive integer. The receiver performs spatial/temporal decoding on signals received by M reception antennas through the channels to obtain reception data without crosstalk between the streams (see, for example, JP-A-2002-44051, hereinafter referred to as Patent Document 1) , where M is a positive integer. Ideally, spatial streams are formed corresponding to a fewer number of transmission and reception antennas (i.e., MIN[N,M]). According to MIMO communication, a transmission capacity can be increased according to the number of antennas, and a communication speed can be improved without increasing frequency bands. Because spatial multiplexing is used, spectrum efficiency is high. MIMO communication uses channel characteristics and is different from a simple transmission/reception adaptive array. For example, IEEE 802.lln, which is a standard extended from IEEE 802.11a/g, specifies an OFDM_MIMO method using OFDM as the primary modulation. Currently, IEEE 802.lln is standardized in Task Group n (TGn) , in which a specification is established based on a specification established in Enhanced Wireless Consortium (EWC) formed in October, 2005. In MIMO communication, in order to spatially divide a spatially multiplexed reception signal y into stream signals x, a channel matrix H is acquired by any method and spatially multiplexed reception signal y needs to be spatially divided into a plurality of original streams using channel matrix H by a predetermined algorithm. Channel matrix H is obtained by allowing a transmitter/receiver to transmit/receive existing training sequence, estimating the channels by a difference between the actually received signal and the existing sequence, and arranging propagation channels in a matrix form according to a combination of transmission and reception antennas. When there are N transmission antennas and M reception antennas, the channel matrix is an MxN (row times column) matrix. Accordingly, the transmitter transmits N training sequence and the receiver acquires channel matrix H using the received training sequence. A method for spatially dividing a reception signal is generally classified into an open loop type method, in which a receiver independently performs spatial division on the basis of channel matrix H, and a closed loop type method, in which a transmitter gives weight to transmission antenna on the basis of channel matrix H to perform adequate beamforming toward a receiver to form an ideal spatial orthogonal channel. For an open loop type MIMO transmission method, there is a zero force (see, for example, A. Benjebbour, H. Murata, and S . Yoshida, "Performance of iterative successive detection algorithm for space-time transmission, " Proc. IEEE VTC Spring, vol. 2, pp. 1287-1291, Rhodes, Greece, May 2001, hereinafter referred to as Non-Patent Document 1) , or a minimum mean square error (MMSE) (see, for example, A. Benjebbour., H. Murata, and S. Yoshida, "Performance comparison of ordered successive receivers for space-time transmission," Proc. IEEE VTC Fall, vol. 4, pp. 2053-2057, Atlantic City, USA, September 2001, hereinafter referred to as Non-Patent Document 2). The open loop type MIMO transmission method is a relatively simple algorithm for obtaining reception weight matrix W for spatially dividing the reception signal from channel matrix H, in which a feedback operation for sharing the channel information between the transmitter and the receiver is omitted, and the transmitter and the receiver independently perform spatial multiplexing transmission. For an ideal closed loop type MIMO transmission method, a singular value decomposition (SVD)-MIMO method using.SVD of channel matrix H is known (see, for example, http://radio3.ee.uec.ac.jp/MIMO(IEICE_TS) .pdf (October 24, 2003), hereinafter referred to as Non-Patent Document 3) . In the SVD-MIMO transmission, a numerical matrix having channel information that uses antenna pairs as elements, that is, a channel information matrix H, is subjected to the singular value decomposition to obtain UDVH. A transmitter uses V in a transmission antenna weight matrix, and transmits a beamformed packet to a receiver. A receiver typically uses (UD)"1 as a reception antenna weight matrix. Here, D is a diagonal matrix having square roots of singular values A,i corresponding to qualities of the spatial streams in diagonal elements (the subscript "i" indicates the i-th spatial stream) . Singular values i are the diagonal elements of diagonal matrix D in ascending order. Power ratio distribution or modulation method allocation is performed according to communication quality represented by the level of singular value with respect to the streams, such that a plurality of spatial orthogonal multiplexed propagation channels, which are logically independent, are realized. The receiver can extract a plurality of original signal sequence without crosstalk, and theoretically accomplish maximum performance. In the closed loop type MIMO communication system, adequate beamforming is performed when the transmitter transmits a packet, but information on the channel information needs to be fed back from the receiver for receiving the packet. For example, EWC HT (High Throughput) MAC (Media Access Control) Specification, Version VI.24, defines two kinds of procedures, "implicit feedback" and "explicit feedback," as the procedures for feeding back the information on the channel matrix between the transmitter and the receiver. For "implicit feedback," the transmitter estimates a backward channel matrix transmitted from the receiver using a training sequence also transmitted from the receiver. A forward channel matrix transmitted from the transmitter to the receiver is computed to perform beamforming under the assumption that bi-directional channel characteristics between the transmitter and the receiver are reciprocal. Calibration of an RF circuit in a communication system is performed, such that the channel characteristics are reciprocal. For "explicit feedback," the receiver estimates a forward channel matrix transmitted from the transmitter using a training sequence also transmitted from the transmitter, and returns a packet including the channel matrix as data to the transmitter. The transmitter performs the beamforming using the received channel matrix. Alternatively, the receiver computes a transmission weight matrix for allowing the transmitter to perform the beamforming from the estimated channel matrix, and returns a packet including the transmission weight matrix as data to the transmitter. For explicit feedback, because the weight matrix is computed on the basis of the estimated forward channel matrix, it may not be assumed that the channels are reciprocal. In view of packet transmission, the transmitter is an initiator and the receiver is a terminator. However, in view of beamforming, the initiator for transmitting the packet is a beamformer and the terminator for receiving the beamformed packet is a beamformee. Communication from the beamformer to the beamformee is referred to as "forward," and communication from the beamformee to the beamformer is referred to as "backward." For example, when an access point (AP) transmits a data frame to a client terminal (STA) as the beamformer, explicit feedback requires that the access point performs beamforming on the basis of channel information transmitted from the client terminal. For explicit feedback, the beamformer can receive explicit feedback of the estimation channel matrix from the beamformee. The feedback format of the estimation channel matrix can generally be classified into two different cases. In one case, a MIMO channel coefficient is sent; while in another case, a transmission weight matrix V for beamforming is computed by the beamformee. The former format is called channel state information (CSI). The beamformer needs to compute the transmission weight matrix V for beamforming by constructing the channel matrix H from received CSI, thereby performing the singular value decomposition. The latter is further classified into a case where transmission weight matrix V for beamforming is sent in an uncompressed format, and a case where transmission weight matrix V for beamforming is sent in a compressed format. According to the explicit feedback, a processing burden for estimating the channel matrix in the beamformer and a processing burden for calculating the transmission weight matrix from the channel matrix are reduced. Fig. 12 shows a frame exchange procedure for transmitting beamforming from the access point to the client terminal by explicit feedback. This procedure is initiated by the access point which sends a sounding packet including a CSI feedback request. The sounding packet includes the training sequence excited by the channel matrix. Accordingly, when the sounding packet is received, the client terminal divides the spatial stream training to estimate channel matrix H and collects the CSI. CSI data is included in the packet as a CSI feedback (CFB) , and returned to the access point. The access point computes the transmission weight matrix for beamforming from received CFB, and multiplies the transmission signal by the transmission weight matrix for beamforming to transmit the beamformed packet to the client terminal. By beamforming, even if the client terminal is located at a place where wireless communication was difficult in the past, the client terminal may still perform wireless communication at a high transmission rate. Subsequently, an operation for performing beamforming according to explicit feedback will be described with reference to Fig. 13. In Fig. 13, a first client terminal STA-A having three antennas is a beamformer, a second client terminal STA-B having two antennas is the beamformee. Feedback is performed based on the CSI format. In the following description or equations, a subscript AB indicates forward transmission from STA-A to STA-B. A numerical subscript corresponds to antenna number of the corresponding client terminal. The training sequence transmitted from the antennas of STA-A is (tABi, tAB2, tAB3) and the signals received by the antennas of STA-A through a channel HAB ace (RAB1 rAB2) .The following equation is obtained. (Equation Removed)where, channel matrix HAB is a 2x3 matrix expressed by equation (2) . Here, hij is a channel characteristics value of the j-th antenna of STA-A to the i-th antenna of STA-B. (Equation Removed)When channel matrix HAB is subjected to singular value decomposition, equation (3) is obtained. Here, UAB is a matrix having an inherent normalized vector of HABHABH, VAB is an inherent normalized vector of HABHHAB, and DAB is a diagonal matrix having a square root of an inherent vector of HABHABH or HABHHAB as the diagonal elements. In addition, UAB and VAB are unitary matrices, namely complex conjugates of transposed matrices become the inverse of the matrices. HAB=UABDABVABH ..-(3) The transmission weight matrix necessary for forming the frame transmitted from STA-A to STA-B is matrix VAB obtained by performing the singular value decomposition with respect to forward channel matrix HAB. When the beamformee receives a sounding packet, the beamformee divides the sounding packet into spatial stream trainings to construct estimation channel matrix HAB. .The CSI composed of MIMO channel coefficients h11, hi2, etc. , which are elements of the channel matrix is collected and fed back to STA-A. If a transmission vector composed of transmission signals of the antennas of STA-A is x, and a reception signal of STA-B is y, the reception signal becomes y=HABx in a case where the beamforming is not performed (un-steered) , but reception signal y becomes equation (4) in a case where the beamforming are performed by transmission weight matrix VAB (steered). (Equation Removed) Accordingly, STA-B can perform spatial division to the original stream by multiplying a reception vector including the reception signals of the antennas by DAB-1 ABH as a reception weight. When beamforming according to the explicit feedback is performed with the CSI format, there is a reduced burden on a process of estimating the channel matrix in the beamformer. However, the terminal, which is the beamformer, computes the transmission weight matrix for beamforming by performing the singular value decomposition, or other calculation methods, with respect to the channel matrix fed back from the beamformee. This is a heavily loaded calculation and the load increases depending on the dimension of the channel matrix. In an example shown in Fig. 13, STA-A includes three antennas (N=3), and STA-B includes two antennas (M=2). Because there are more antennas in STA-A than in STA-B, no problem is caused in the processing capability for beamforming. This is because STA-A is designed to include the processing capability corresponding to N of its own streams; and an NxM channel matrix is constructed on the basis of the CSI fed back from the beamformee to perform computation of the matrix for beamforming on the basis of the channel matrix. However, for N