SIGNAL PROCESSING UNIT EMPLOYING A BLIND CHANNEL EaSTIMATZON P,LGORITHM AND METHOD OF OPERATING A KECGNEF. APPARATUS Embodiments of the invention refer to p, signal processing unit employing diversity 5 combining. Other embodiments refer to a method of operating a receivel- sppnsat~u1:tf an electronic device. Electronic devices may be provided with multiple receive antennas to re.dize diverairy combining, for example EGC (equal gain combining), MRC (maxim>-in1 ratio 10 combining), or an SC (selection combining) approach like SSC (sv~itch.e .nd ~ t a y combining). Diversity combining relies on channel state information describing the transmission properties of each transmission channel. Knowledge of the amplitude and/or phase information of each transmission c-hanael i3 obtained from demodulated receive signals, for exampla by comparing training symbv1.s coiitained .iu 15 the receive si$ne.l.ls with reference ififormation available at the receiver side, for example reference signals. The object of the present invention is providing a cost-effective signal processing l-uiit employing diversity combining. This object is achitved kvith the su,bjc~:t-matter of the 20 independent claims. Further embodin~cllts are specified in t.he dependent c1ai.m~. Details and advantages of the invention will become more apparent frog? the following description of embodiment8 in connection with the acc.ompanying drawings. ERlEF DESCRIPTION OF THE DRAWINGS 25 Figure 1 is a schenlatic block diagram of a receiver apparatus employing diversity combining for discussing background information uuable for undel-standing the invention. 1 clr ANN & PAKTNER 1 SONY Corporati~n 63613 Figure 2A is 4 schematic block diagram of an electronic device including a plurality of tuner circuits a&olding to an embodiment related to wireless. communications systeins. 5 Figure 2'8 is 4 schematic block diagram of sn electronic device including s. plurality of tuner circuits according to an embodiment related to wired commlnications systems. Figure 3A is a schematic block diagram of a receiver apparatus employing diversity # . 10 combining in accordance wilh an embodiment. Fig-ure 3B is a schematic block diagram showing details of an embodimeat of the signal proceasing unit of the receiver apparatus of Figure 3A. 15 Figure 3C is a schematic block diagram showing details of an embodilllent of the. estimator unit of the signal processing unit of Figure 3E; in accordance with a11 embodiment related to a time-domain oriented definition of an enor signal. Figure 3D is a schematic block diagram showing details of an em,bodiment of the 20 filter unit of the signal processing unit of Figure 3B. Bigure 3E is a schematic block diagram showing details of &II embodiment of the estimator unit of the signal processi~lg unit of Figure 3B in ~ccordancz with an embodiment related to a frequency-domain orient& definition of an error signal. 25 Mgure 3F is a schematic block diagr8.111 showing details of another embodirrlent of the. estimator unit of the signal processing unit of Figure 3R in accosclat-LCiner unit. The signal grocessing unit 100 may output the demodulated signal to a cgntrol unit 29 0. 5 The control unit 290 may control an output device 29.1. to output infol~llation encoded in the demodulated signal. For example, the ~ u t pu~ni~t t2 94 nlay hp n screen displaying a television program encoded in the transmit signal s(t). According to another embodiment, the output unit 294 outputs audio information cr~codtd in 10 the transmit signal s(t). The electronic device 200 nlay further include an input unit 296, for example a keypad or a sensor array applied to the output 1ir3it 294 for allowing a user to control the electronic device 200. Figure 2B refers to an embodiment of an electronic device 200 for a wired 15 mmmunication system using a siagle-input- multiple- output (SIMO) or 8. rnt~ltiple- , - input-multiple output (MIMCJ) approach, for exmple DSL (digital subscriber I.ine) or PLC (power line comn:unications). The receiving elements 205 may be terminals of a connector block receiving, by \nay of example, one, two or three life wires, s ne.utra1 wire and/or an earth wire of a power line network. Tu=er circuits 208 may tune in to 20 the same carrier frequency and supply receive signals ri(t) to s signal processing unit 100. The signal processing unit 100 performs a blind chu.n:lel estimation for obtaining chflanel information. The signal processing unit 100 may apply diversity combining to obtain a coarse estimation of the transnlitted signal on the basis of the results of the blind channel estimat.ion, may or may not pre-equalize the con~bined 25 signal on the ba.sis of information obtained by the blind channel estimation, demodulates the combined or pre-equalized signal, and outputs the dernodulatccl signal to a Control ~ m i t2 90. The control unit 290 may prol-euo the clemodulated signal and an interface unit 298 may output the processed signal, 30 Figure 3A shows details of a signal processing unit 100 being iptkgrated in a receiver apparatus of an electronic device 200, e, g, in a televisio~l receiver- The receiver apparatcs includes receiving elements 205 and tuner circuits 205'. The s i g ~ a l processor -mit 100 may include a plurality of A/D converter units 110, wherein each MULLER H O F F M ~ NR ?.w~\!E?. - - 10 - .$, SONS Corporation 63612 17.02.20 11' I AID converter unit 110 samples the respective analogue receive aigns.1 ri(t) at a predefined sample rate 1/TA. From the sampled receive signals ri(n.TA), an estimator h 7 unit 120 derives the estimated impulse response /I, for each ~ a ~ ~ i prleccediv e signal C ri(n.TA), wherein the estimator nnit 120 may adapt an iterative blind channel 5 estimation algorithm according to a sparseness rncssure obtained by the estimateu. A combining unit 130 combines the sampled receive signa,lls ri(n.T.4) or signals derived therefrom on the basis of combining coefficients derived from the estimated impulse - responses hi to obtain one combined receive signal y(n.T~)w, hich is a11 ~ s t i r ~ l aotf ta transmit signal s(t) transmitted via the plura.lity of transmission channels. 10 One or more demodulator units 180 of different types DEMA, DEMB, ... 1li8Y synchronize and derxodulate the combined receive signal y ( n . T ~b)y rn~tlti~lgintgh e combined receive signal y(n.TA) with an analyzing signal hawing ri dcmctdulator frequency. One or more of the demodulator units 150 nlay use the comhined receive 15 signal y(n.TA) in the time domain. According to an embodiment, at least one. of the. demodulator units 180 uses a frequency domain representstioa of the cmmbined receive signal y(n-TA). Then a transforrnatiot~ unit 17'0 may be provided in a, signal path between the combining unit 130 and the demodulator units 180 using the frequency'domain representation. The transformation unit may ~.ppJya DFT (discrete 20 Fourier transforrr.) to the combined receive signal to obtain a corrtspoi~ding description of the combined receive signal in the frequency domain. One or more of the demodulator units I80 may be c.onfigured to dezr~odulate analogue television broadcast signals foll~wing different television hroadcast 25 standards, other demodulator units 180 may be adapted to demodulate digital television broadcast signals, digital radio broadcast signals sr analogk~e radio broadcast signals. According to an embodiment, at least one of thc de~nodulatur units 180 may include a quadrature demodulator, or P,B OFnhl (orthogorlal ft equency division multiplexing) demodulator. One or more of the demodulator units 180 may 30 further modify the demodulated receive signal using ~nformatiore~n coded in the receive signal. The demodulator units 180 may include bit error corre.c.tion unite. G SONY Corporation 63612 using bit error detection and correction approaches to correct bit errors in tkre digit4 data stream. As regards the estimator unit 120, the idea may be a blind diversity combining S approach deriving the combining coefficients Zi for the receive signal ri(t) from a blind channel estinlation of the respective channel impulse response hi(t), whereis a blind channel estimation algorithm is modified according to sparseness information about the estimation results, wherein the sparseness information conta,ins information abont a change of a sparseness measure. kemwk~bly, though blind 10 channel estimation algorithms contain at least a scalar ambiguity in their estimates, dhersity combining cen still be applied at a profit: As the i.nventors could show, diversity ctimbining (even MRC) is robust against the kind of scalir ambiguity introduced by the blind channel estimation algorithm. Hence, other than in, conventional approaches where determination of the chanrrel impulse response 15 depends on knowledge about the structure of the transmitted signal. (e. g. about whether the signal is an OFDM signal etc.) and/or a'oout a training sequence in the transmitted signal and typically requires esact time, freq~lency and sanlple clock synchronization before reference or training symbols can be ide.ntided and used for channel estimation, blind diversity combining as -pro?oszd herein neither r e i ~ ~ ~ i r e s 20 frequency nor sample clock synchronization before the ccrnbinin.g step. i Instead, blind diversity combining exploits cross-rela,tion statistic.8 betsveen the receive signals and uses not-demodulated instances of the receive signals. The: blind diversity combining may be based on higher order statistics cross rdations. 25 According To an embodiment, the hlind diversity comhlning uses second-order statistics crose relations. According to further cmbodiinentr, the blind diversity combining does not rely on knowledge on the type of rnodulotiun applied t u obtain the transmit signal and on the. type of dcnodulaiio~x re.quii-ed to demodulate the receive signal. Where conventional approaches which consider individual channel 30 transfer- functions have to instantiate a demodulator unit emph3yiflg demodulation and synchronization for cach of the receive paths, 'the blirid char~nel estimation requires this portion of the dernodulalo; anit ozly once for the combined receive signal. Where systems contain several types qf demodulators, each ty~led edicated to u ? - r c c - ; u w I J PIUI ler nor ~mann ,,,,A N * - t, + A ! - HY I I I ~ ~ L. 8 . . 1 1 . 0 .L 4 " .l.-M+tbL S) SoNY Corporation 63612 - 17.oz.2.012 another communication system technology or com~lunication standard, blind diversity combining allows ihstan~iatiag only one single dernoriulattor for ea.ch co~munications ystem or communication standard respectivrely. As a rc.sult, 111 a MIMO (multiple-input-multiple-output) or SIMO (sitigle-input-multiple-output) 5 communication system blind diversity combining reduces costs for diversity receivers significantly. Figure 3B shows details of the estimator unit 120 and the colnbining unit 130 of a signal processor unit 100 according-to an embodiment applyi,rg a bli.lzci channel 10 estimation based on a cost fanction defined in the time donlain in combination with MRC. The estimator unit 120 receives the sampled receive signals ri(nTA). hc.cording to an embociment, the estimator unit 120 applies the MCLMS (multi-channel least mean squares) npproach. 15 The MCLMS approach relies on the observation that in the absence of noise for each pair of receive signals equation (4) liolds: 20 The observed deviation from this assunlption is cor.sidered to represet~t a.n error signal au regards the cross-relation between the receive signals i a ~ j~. Fdro m the error signals of all pairs of sampled receive sig.n-a ls ri[n.TA) cost fanctior~is defined. + h e cost function includes a cross-relation matrix containing infornlation ahour the Gross relations between dl pairs of receive signals. An iterative approach for ~1.1 25 estimation of the channel. impulse responses hi(11aTA) obtains updated va'lldzs hi((a+l).TA)o n the basis of previously obtained values hi((a)*'TA)t,h e updated error signals and thc updated cross-relation matrix. According to ancther emhodimer~t, a L4CN (multi-channel Newton) dilgorithrr. may be applied to acce1erat.e the convergei1c.e of the algorithm. In accordance with further embodimento,, the estimator unit. 120 30 instantiates a plurality of different blind channel estimation a2gorit.h~n.s and m3.y select the current blind chanseJ estimation algorithm according to e.n internal stale or in response t.o a first selection signal sell. The cst,imntor xrllit 12Q ~ulltinuously - - - - - " V . 2 -I" . Y ' * Y I'IV.I IVI IIVI I IIICh, ,I l 1 1 1 1 5 1 1 . . J T L V J TUJLIU.JJ W. C11U.f UJL 1) SONT Corporation , 63612 outputs the results of the estimation, e.- g. the esti~natedc hannel irnpulsp, responses h., to a control 'unit 132. I d Based on the estimated impulse responses hi and t.he predefined or cl,lrrently 5 selected combining scheme, the control unit 132 msy continuously update filter coefficients of filter unita 134 provided in the signal path of the sampled receive signals. The confro1 unit 132 n a y be programmable and may switch l:\etwccn bIRC, EGC and SC upon an appropriate second sclectiorl signal se12 or in'resgo:lee to a. change of an .internal state. According to an embodiment prnvidiilg SC, the control 10 unit 132 configures a filter unit 134 arrsnged in the signal path of the re.ceive signal which has been' identified as t.he leaat disturbed receive signal by the estimated A - impulse responses hi to let the, least disturbed receive signal pass, whcreae all other filter units 134 are configured to block th= other receive signals. 15 According to an embodiment applying hlVIRC, the control unit 132 derives combining coefficients that configure the filter units 134 to represent matched filters which match with the respective estimated trailsmission channel rcsptctivelp. Each filter unit 134 convolves the respective receive signal with the respective matched filter function. 2 0 A superposing bnit 136 adds up or superposes the filtered or weighted receive .signals to generate the combined receive signal y(n-TA) representing an estimation of the transmit signal. 25 The signal processing unit 130 may further comprise a pre-equalizer unit 135 which equalizes the c.omhioed receive signal and which outputs a pre-equalized si.gno.1 z(n.TA). The pre-eqoalizer unit 136 may use inforl-eation obtained froin the blind channel estimation algorithm. Applying blirld diversity selecti,oa, the receive signal znay have approximately twice t.he length of the loligest chani~c! 30 impulse response. This would directly increase the rsquired order of an cqudilizcr that is effective in the following demodulator stage. In OFDM systems a guard interval violation could occur. According to an embodimei~t, the pre-equalizer unit 138 applies a delay spread minimization algorithm. According 16 another embodiment, the prc-equalizer unit 138 ia designed to mitigate fast time-vzaiant channel effects svch as inter-carrikr interference in a subsequent OFDM demodulator.' Pre-equalizing is performed before de~nodulatinn of the con1bine.d 5 receive signal. Pre-equalizing may be performed before or after Fourier transforming the combined receive signal into the frequency domain. According to an embodiment providing the MCLMS approach, the estimator unit 120 estimates channel impulse responses using only sec,ond-prder statistics of the 10 receive signals ri(l) in the time domain. F.or example, t.he estimator unit 120 applies a cross-relstion approach minimizing an error f~uiction derived from c?.ross-relation information concerning pairs 5 (t) , r. (t) of receive signals reupectivrly. In the 1 absence of noise, equation (1) can be rewri.tten as indicated in equation (5): Equation (6) reflects eqhation (5) in vector form: (6) (n) L (n)= F1 I( n)i i( n) 20 With Z defining the maximum length (channel order) of the cha~lael impulse, - responses, equation (6a) gives a cbennel impulse response vector hi(l-1) for chaflfiel i; 2.5 In the presence of noise, the terms on both sides of equatiori (6) iiiffel- from each other. According KO an embodiment, the difference defines an error signal ~ ~ ( 7 1 ) between the two channels i and j: . . . . - f l ! ~ - b k k - ~ u l - I GAYII NI .-. . ..- L. AC 1L ,I.RQ w.22 L I.3 M-. HV~~MANN PARTNER SONY Corpora:ion 63612 Prom the error signals eii(n), a cost function may be dcriv~dth at considers all error signals eg(n). Equation (8) gives a.n example for a cost function in accordance. with an ernbcdiment that is based on all error signals ei(n) between each pair of the receive signds: Other embodiments may rely on differently defined error signals or on other .C.. derivatives of the error signals ey(n). An estimate h of channel impulse response - 10 -vectors h aims at minimizing the cost function J(n) according to equatioh (9): (9) h = arg minS E {Y (n)} subjest to 1Fll=1, wherein 11h1'1 represents the magnitude norm of the eatimstccl 15 channel i a p d s e response vectors. Calculation of the estimated chanuel impulse - - response vectors h may rely on the observation that the vector 12. of ckai~neilr np~dsc responses lies in the null space of a rnatrix R whose entries include autocorrelrd tl' on su~dc ross-correlation information about the receive signals t;(t) as given in equation The matrix R ie given by the matbema.tica1 ekpcctstion values R,,, of two receive signals ri(t) and r .(t) , resptc~ively: J SONY Cargoration 63612 17.02.2012 - In equation (lab) E{ ...I represents the mathematical expectation operator. According to an embodinlent, the estimator unit 120 minimizes the cost func.tior by updating the estimated channel impulse responSt vector adaptively according to equatiorl (1 1): In equa:ion (111, p represents a. predP;fn.ed positive step size parameter. The MCLMS algorithm converges stea.dily to the desired solation, According to another embodimeat, the MCN (multi-channel Newton) algorithm is applied to accelerate the 15 convergence. Figure 3C schematically illustrates an embodiment of the estimator Unit 120 instantiating th.e algorithm accarding to eqoations (5) to (1 1). A correlatio~i'u nit 121 derives autocorrelation and cross-correlation information from the receive signds 2 3 r(tj and provides information describing a cross-correlation matrix approxinlation .i?. - A storage unit 129 stores the estimared channel irupulse response vectors IT(&) obtained from the preceding iteration step. An iteration unit 125 determines the * updated estimated channel impulse response vector i(n+l) from the cross-. -. correlation matrix approximation R, the step-size parameter 11 and the pre.vioc.r:: n d 25 estimated channel inlpulse response vector k(n). Since no information about s(t) lu uued, blind diversity conlbining is independent from the applied transmission standard and can be applied to different transmission systems and standards. Blind diversity combining works with both single-carrier and 30 multi-carrier signals and does not require a redesign of existing demodulators. - 17 - SONY Corporation 63612 17-02,2012 Since the estimation unit 120 does not synchronize in time with the receive signals, the receive signal may bo sampled at a sample rate (sample frequency) tha,t is higher than an upper frequency limit of the. receive signal. For example, the sample rate is twice the upper frequency limit of the receive signal. As a result, the eampled receive 5 signal rnay also contain portions from outside the receive signd (out-of-banci energy). The out-of-band energy increases the number of possibilities fur a channel estimate and may lead to a misconvergence of the estimate. According to an embodiment, the estimator unit 120 may include a detecting -i~nit 10 12Za and a power limiting unit 128b. The detecting unit 128a mag detect a misconvergence of the iteration algorithm or a criterion that puteatially may result in a misconvergence. For example the detecting unit 1283 may evaluate differences between two or more consecutive estimates for the channel vectors. When, at a certain iteration step, the differences between two or more consecutive estimates for 15 the channel vectors exceed a predetermined thieshold, the detectine unit 1283 may output a controi signal indica.ting a misconvergence to tKe power limiting unit, 128h. According to an embodiment, the detecting unit 1'383 determines the out-of-band signal energy of the estbated channel vectors and outputs a co~ltrol signal indicating a potential n~isconvergenceif the de.termined out-of-hand energy exceeds a 20 predeterr~ined threshold. The evaluation of the out-of-band e.ne1-gy may be pel-formed adapcively by uuiag varying thresholds depending on a previvusly applied threshold and t3e currently estimated channel vectors. 14 response to the cantrol signal, the power limiting unit 129b attenuates higher 25 frequency portions in the estimated channel vectors and the itc.i-ation al~orithmu ses the attenuated estimated channel vectors in the nest iteration step. The higher frequency portions are ir~a frequency range between th-e upper frequency limit of the receive signal and the sample frequency. According to an ernbodimcllt-, the power limitation or clipping rnry be performed iteratively. The power lirrliting Uhit i28b ~haY 30 release or successively decrease the attrl~uationo.f the, higher freqt1e.n~p~o rtions when the detecting unit 128a detects a more convergent behaviour of the iteration alga~ithma nd outputs a corresponding control signal. - SoNY Corporation 63612 Tht attenuation of the higher frequency portions may start for frequencies just below the sample frequency and may successively proceed ta lower frequet1c.ie.s. Accordizlg to another embodiment, first a small attermation js applied to early dl 5 frequency components between the upper frequency edge of the receive signal snd the sample frequency and the degree of attenuation is gradually increased. cither embodiments may combine both approaches. According KO an embodiment, the power limiting unit 128b filters the estimated 10 channel vectors by limiting the maximum affordable power for frequency band edge areas of the estimated channel vectors. The power limiting unit 128b may use a DFT or a kind of filtering ro obtain the required information and to limit the out-of-b.ad energy. The pourer limiting unit l28b may be provided in a signd path between the output of the iterst.ion unit 125 and the input af the st.orage vnit 129 or in a signal 15 path behveen the output of the storage unit 129 and an input of the iteration unit 125. Figure 3D refers to details of the filter units 134. Each filter unit 134 may cunlpi-ise delay units 134a for providi~lg delayed instances of the respective receive signal 20 ri(t,n-TA) delayed by multiples of TA. P, configurable .weightir~gu nit 134b is as~ignerl to each delayed insta.nce, which multiplies the respective instance of the concerned receive signal ri(t,n-TA) with the corresponding ser of combinir~g coefficients ci(t,nTA). The outpnt signals of the weighting 'units 134b are combined Qr superposed using summation units 134c to generate the filtered signal yi(t,n.'l'A). 25 Other cnlbodirnents of the estimatar unit 120 may provide an iterative estimation of - chann.el vectors hi( m) in the frequency domain (chanlel tran~ferf ur~cti.oi~sw),i th the. index (i) identifying the channei number. For example, the estimator unit 120 illustrated in Figure 3E refers to an einhoc;lme.nt. 30 providing the MCFLMS (multi-channel frequency least mean square) .algorithm. - MCFL4MS derives frequency-domain block error sequence vectors -e.' J. (ri) for all M channels on the basis of an overlap-$avo or overlap-add terhn.i.q~~crs ed far frequency-domain filtering. For a current block rn of 2L sample, values, MCFLMS v3-rLL-Lulj-ula .u: LU Mu 1 1 er Hot i mann PAX 1:. :f4k %Y I B Y l l I J 3 3 M- & PAEWER ib SONY Corporatiox 63612 coflstructs a frequency-domain mean square error critcriol? nncllogous!y to the abovrdiscussed time-domain approach. Then an iteration algorithm updates the frequency-domain channel vectors hi (m) for the next block (m+ 1) - 5 .For applying the overlap-save tecb.nique, first a vector ylj(nz) (i, j = 1,2, ..., M) of length 2L may be defined that results from the circular cot~volution of chc receive - signal vectors 3 and the estimated channel impulse response vectors A . as given by I equations (12),( 131,( 14),a nd (15): I 15 (14) c, (m) = ( - 1 ) ... );(nzL+lj With equation (12),t he last I, values from y..(rnt) to Sj..(~td-~.L.-alr)c identical. to LJ 11 20 the results of a linear convolution as given by equation (1 6): with 25 ?r1)-~I:C-~Ul3-U-I.C' WIVI ar~rTrnann PHA Li:. 03 T G Y L l U j i -. s 2 i . _ l ~ m < H- - 2 (11 - SOWY Corporation 63612 17.02.20 12 - In equation (16), hj(n1) is the estimated channel vector in the frtc1uency donlain for the j-th channel. The W matrices are used to realitt the zero padding anil ~nvsking 5 for the overlap-save technique. In particular, utilizing the. overlap-save technique, the input data blocks are overlapped by L, points. For each bloclc of length L,, whe.re 2L data inputs are available, the first L circu1.w convolution results M-e discarded and the last L convolution results are retained as outpats. Equation (16) call be used to define a time-domain block error signal vector E-(r?t) between the i-tti and the j-th 9 10 channels for the m-th block on tke bsvis of the respective Cri matrices: Iri equatiofi.(l7) Cq and C are circulmt matrices. The MCPT,MS a.pproac11 exploits 15 the fa,ct that C, can be decomposed in a diagonal matrix DG in the f r t - q ~ e n ~ y '1 domain using equation (IS), with FZLr, beiag the Fourier tra~lsforlnm atrix ~ o sfiz e. 20 With equation (18) it can be shown that freqx~ency-domain b1oc.k error sequence vectors -z41 -(m) can be used in the iteration step instead of the time-do~nairb~lo ck: error signal vectors zii(nl.). To this purpose., the frequency-domain block error sequecct vectors -sV- .(rn) are defined as the Fourier transformed of the time-clomain 25 hloclc error signal vectors q(m) by nlultiplying the time-domain block eilSor signal vcctors G(m)w ith a Fourier matrix of size L x L ~lcc.ordingt o eq~~atio(1n9 ): M m IIC+i%MW S PA- - 21. - I + SONY Corporatiah 63612 17.02.20-1- ?, .--.. Inserting equation (17) in equation (19) and then applying equation (18) gives equation (20): 5 Due t o the nature of the diagonal matrix Dq the computational burden to obtain the 7 frequency-domain block error sequence vectors eg(m) is low compared to that required to obtain the time-domain block error signal vectors ii;,.(nz), such that in tb,e 11 frequency domain a fewer n.urnber of multiplications replaces the higher number of 10 multiplications in the computation all^ intensive ~onvolutiono f the i-th receive sigrld r, (n) and the impulse response hj(n) of the j-th channel estimation, which is required for the calc~lation of the time-domain block error signal vectors i?,.(rn.) !I according ro equation (7). Then a frequency-domain mean square. error criterion Ec!r the m-th block is constructed ~nalogously t.o tbe time-domain approach. From the 15 error criterion, the frequency-domain LMS algorithm cpdates the freque~icy-domain channel vectors for the (m+I)-th bldckr from rhe frequency-domain channel vectors of the m-th block according to equations (Zl), (22,), alld (23): 35 In Figure 3E, the estimator unit 120 includes register units 122 20 provide, for each receive path, 21, sampled values of the receive. signa.1~i s o.ppropriate form to discrete Fourier tramform units 123. The Fourie; transform units 123 o ~ t p u tt l ~ rtc r;pect?$e diagonal rpatrices to an error calculator unit 124. The error calculator U r i t 124 uses overlap-save approach to determine a frequently-dolnaix block t.rrur sequeIlce 30 vector uving the Diagonal mat.rices and the previously obtained estimated frequency SONY Corporation 69612 17 (I>.ZIJ~~ . -- domain channel vector. The error calculator unit 124 outputs the frequency-domain block error sequence vector fa an iteration unit 125. The iteratimi unit 125 updates the estimated frequency-domain channel vectors on the basis of the previouvly obtained estimated frequency-domain channel vectors, the ~ ~ p d a t afdr equency- 5 domain block error sequence vector, the Diagonal matrices and a predefined step size parameter pp, A storage unit 129 stores the updated estimated frequency-dorr~airl - channel vectors h for the next iteratio~xs tep. An i~~verstrea ilsformation i~nit1 26 applies an inverse discrete Fourier transform to obtain the l~pdated time-damaiu - channel impulse response vectors hi . According to other embodimerlrs, the 10 estimator unit 120 includes Curther sub-units to perform a MCNRLMS [multi-ch~nnel normalized frequency least mean square) algorithm. According to an embodiment, the estimator unit 120 comprises a sparseness evaluation unit 127 and an adjustment unit 125% The sparseness evalus.tiou unit 15 127 estimates a sparseness measure gn) on the basis sf the. estimated channel -A impulse response sectors h. The adjustment unit 124a modifies the estimated -* channel impulse response vecrors h on the basis of the sparseness measure in s way that channel sparseness of the estimated channels is increased. 20 In general, sparseness measures give the degree of sparseness for an impulse response, wherein impulse responses that concentrate in few coefficients are $parser and less dispersive than impulse responses that distribute over more coef-fiicier,ts. Usually a channel is the sparser the fewer n ~ n - ~chtcrirt~ne l coefficients t5.e cha~iilc'l has. The sparseness measure may be any sparseness zneasure publisilecrl in the 25 cohtext of transmission channel spareeaess. According to nn embodiment, the spmseness rneasure zn) iu given by equation (24): . . PAX N ' . 't!? . . I U & H H .- . r r m I 'AllI le. h, D2:j;u22 I W ~ KHO FFMANN & PARTNER - ,--I3 ,.. - SONY Corporation 63613 17.02.2012 with 0 5 x r z ) 11. A low value for gn) indicates a nearly dispersive chailnel alld a value close to 1 indicates a sparae channel. With the sparsenes9 mcasurc. as defined in equation (24) blind diversity combining schemes yield mare reliable results. 5 The sparseness evaluation unit 127 rnay use tine-dornnin. inlpul~e responses obtained by inverse Fourier transformation of the delayed (stored) estimated - frequency-domain channel vectors. According to rhe illostratcd err~bodimcint., a second storage unit 127c delays or stores the updated time-domain channel ini~uIsc response vectors obtained bp the inverse transformation unit 126 ~inds upplies them 10 to the sparseness evaluation unit 127c. - The adjustment unit 125a adjusts the estimabed frequency-dornai~c hannel vr.i-.tors on the basis of the obtained sparseness measvre. For txample, the adjustnlent uriit 125a reduces the estimated frequency-domain channel vectors within the iteration 15 loop by an amount defined by an adjustment term derived from the obtained sparseness measure. The adjuetnient unit 12% may modify the clpdated estiniated frequency-dumain channel vectors on the basis of an adjustmmt term that inweases the sparse-ness of 20 the estirPat.ion with ongoing iterations, rhereb'y exploiting the fact that for man17 applications real transnlission channels appear sparser than their estimates. For example, in real broadcast transmission obstacles refle.c.t the tru.ncsrniseion sigilals. As a result, various delayed versions of the transmission signal superpose each otker at the receiver side. Each reflection of the real transmission signal corresponds tcl a. 25 high coefficient value in the channel impulse respondt at a time point given by the position of the ibstaclc. It can be observed that impulse responses for real transmission channels aften cor~tain only few strc1n.g coefficient2 resulting from reflections, whereas the impulse respolises are flat and near ztro in the rest. 30 The adjustment term may or may not consider previous channel estimationu. According to an embodiment, the adjcstment tern1 considers both the sparseness measure and at least the previously estimates channel impulse response vectors. An example is given in equations (25), (25a) and (25b),w herein the adju,r t.n xent term co~itainsa n adjustment parameter ,$ derived from the spsrsene.ss trrleas.i~re$ 01). 10 In general, the zdjustment unit 1253 may modify the updated estin7.s-ted freq~~encydomain channel vectors such that for each transrnissioll chn.nnel the sun1 of the chaznel impulse response coefficients is nli~lilnized e,nd at the same time -an. energy represented by the respective transmission channel impulse rtspocse is m%intaiaed. 15 The adjustment patameter ;Zk is derived from the current sparseness measure su.ch that an unwanted error floor car1 be avoided. The derivation of the adjustlnent parameter Ak from the sparseness measure n a y depend on various constraints such as the assumed sparseness of the red transn~issionc hannel, 20 According to an embodiment, on the basis of the sparseness meavure a differentiator unit 127L may provide the adjustment parameter ak such that the adjustment term in equation (25) is relevant only as long as the algorithm has not s~lfficiently coilvkrged to the desired solutiop. For example, the acljustrrlent parameter Ak 1s 25 selected to represent nn approximation of the differential of the sparsenass measure with respect to the iterations- According to an emhodiment, the difie.refitiatar unit 127b sets the sdjustment parameter Ak according to equation (26): (2 6) jlk (n) = 3" Ign) - gn - 111 30 M- . HUI-~MH-R SONY Corporatioa 63612 Usually, at the beginning of the estimation the sparsehess measure varies by a large value until it converges to a ~ e r t a i nva lue. Then, in the case of a static channel, the sparseness measure stays nearly constant. For time-vslying channels, th.e sparseness measure varies only by a small mount aiter the initial phase. Hence, 5 eqnation (26) ensures that the adjustment term is active only as long as it is usef~ll during estimation. For example, for a static channel the weight is colltinuously decreased, whereas each channel change triggers an increase of the weight.. The constant y may be selected according to application constraints. 10 According to another embodiment, the adjustment parameter ak is defined on the basis of an averaged time derivative of the s2arseness measure. Fur example, the adjustmeat parameter /Zk is defited as in equations (27a), (27b): In equation (27b) a forgetting factor gives the weight of the current ac!justment parameter /Zk with regard to that of the previous iteration and allows control of the 20 short-time averaging in a way to remove the influence of noise. The forgetting factor p can be set according to the application or can be. adapted to current noise conditions. While Figure 3E shows an example of the configuration of the sub-units of the 25 estimator unit 120, other embodiments may pravide other cor~figurations. For example, the embodiment of Fimrr 3F provides an inverse trs.nsformation unit 126 providing a,s ixverse discrete,Fourier at the output of the ittra,tion unit 1125 and. supplies the time-domain impulse response vectors output by the iteration unit 125 to the sparseness evaluation unit 127. The modificatioi~ takes place in the time- 30 domain. An additional transformation unit 123 applies a discrete. Fourier trarisforrn onit 123 or1 the modified impulse response vectors to obtain modified frequencydomain channel vectors. The frequency-domain channel vectors are supplied f ~trhe ~ ~ - P c c - L u I s - Y I I ! - u : I.1 lvluller Hottmann FAX 1:. : t 4 L H!i Iki!-II-]Iljj <. 11'2 1 !I[-,,; .'. cn2 I . , V-.b - h&J N FI- & PARTNER r) - SOMY Corporation 63612 next iteration step. With the embodiment of Pfgure 3F each itexration step is bn.sed on the sparseness measure of the channel vvectors of the preceding iteration step. The diagram in the upper half of Figure 4 shows sn estimated channel iinp~rlue 5 response for an algorithm that does not consider sparseness. The diagram in the lower half of Figure 4 visualizes the effect of an algorithm according to the embodiments airning &t increasing the sparseness of the channel estimation -andcr the assumption that the real transmission channel shows high spal-srness. The diagrams axe for illustrative purposes only and do not reprod11c.e actuai 10 measurement or simulation results. I Since the estimation unit 120 does not synchronizc in time with the receive signdils, the receive signal. may be sampled at a sample rate (sample frequency) that is higher than an upper frequency limit of the receive signal, i.e. the receive signal iis 15 oversampled. For example, iht sample rate is twice the upper fi-equc~~cliyrn it of the receive signal. As a result, the sampled receive signal may also contain portions from outside the receive signal representing out-of-band enera. The out-of-band energy iis assigned to a freq-aency range between the upper frequency limit of the receive signal an2 the sample frequency. The out-of-band energy increases the cumber of 20 possibilities for a channel estimate and may kad to a misconvergcncc of the tstimale. Applying an iteration algorithm that modifies the cost fur~ctiorlr - ~ ~ i ra~ g term that iu continously adapted to the channel behaviour reduces the number of possibilities for the channel estimates and results in a more c.onverge.nt behaviour of the iteration algorithm. 25 According to an embodiment, the estimator ini it 120 nlay further iilclude s detecting wit l2Sa and a power limiting unit l28b. The detecting unit 128a IIIELd~e tect a misconvergence of the iteration algorithm or a cl-iterion that potentially may result. ir-1 R. misconvergence. For =ample the detecting unit 122a may evaluate cliffcre~~cee 30 between two or more consecutive estimates far the channel vectors. When at a certain iteration step the differences between two or more consecutive estinlates for the channel vectors exceed s predetermined threshold, the detecting unit 1233 may output a control signal indicating a miuconvergence to the power lirnitil~gil llit 12Sb. SONY Corporatiu~z 63612 17.02.20 12 .- According to an embodiment, the detecting unit 128a determines the out.-of-hand - signal energy tif :he estimated channel vectors and o~.ltputs a control signal indicating a potential rnisconvergence if the determined out-of-band energy exceeds a predetermined threshold. The evaluation of the out-of-band energy may Ise performed 5 adaptively by using varying thresholds depending on ~c previo-~~salyp plied threshold and the currentlqb estimated channel vectors. 111 reSpOhSc to the control signal, the power limiting unit 128b may attenuate the out-of-band frequency portions is the estimated channel vectors a,nd the iteration 10. algorithm continues with the attenuated estin~ated channel vectors. in the 11ext iteration step. The out-of-band frequency portions are portions in a frequency range hetween the upper frequency limit of the receive signal and the sample freq~1enc.y. According to an embodiment, the power limitation or clipping may he performed iteratively. 'the powex limiting unit 128b may release or eucr,es~ively decrease t.he 15 attenus.tion of the higher freqkency portions when the dct.ecting unit l28a detects a more convergent behaviour of the iterati011 algorithm. and o-fitputs a carresponding control signal. The attenuation of the higher frequency portio~ls may start for freqliallcies just 20 helow the sample frequei~cy and may succesuively proceed to lower frequc~l~ies. According to another embodiment, first a small attenuation is applied to nearly 311 frequency components between the upper freque~cye dge of rhc. receive signal arld the sample freqxency and the degree of attenuation is gradually incres.scli. Other embodiments may combine bath agproaches. Tht power limitkig unit 128b may be 25 provided in a. signal path between an output of the iteration unit 1'2.5 a.nd an illput cf the storage unit 129 or in a signal parh between an output of the storage unit 729 and inputs of tht iteration and error calculator units 125, 1'24. According to an embodiment, the detecting unit 128a dzterrniaes an out-of-band 30 energy of the estimated frequency-domain channel vectors. When the deterlni,ned energy exceeds a predefined threshold, the detecting unit 128a 1:lutputs a cor~trol signal to the power limiting unit 128b. In respoilst to the contra1 signal the power limiting unit L28b l i ~ ~ i t s / c l i p s powel- used for the estimate in the. oversampled H P SONY Corporation 63613 areas in order to control the estitnation process in these ereas to obtain powerlimited estimated channel vectors, Applying a power limitation in t h e out-of-band areas reduces the number oi possibilities for the channel estirnatea and resu!ts in a more convergent behaviour of the iteration algorithm. According to a further embodiment, the estimator unit 120 inclildes all SNR unit 125b thdt outputs SNR information descriptive for current signal-to-noise r3.ti.o values of the channels. The inventors could observe a linear rel~tionshipb e tbiu'eent he cost function in logarithmic scale and the SNR value for small SSNR values up to 10 about 30dB when the cost function has converged. Since wireless conlr~~unications provides low transmission power a?d deals with missing line of sight, fading and other distortions, typically a low SKR below 30dB is observed at the receiver side. Hence the linearity between the converged cost function and the SN2 as given in c yuat,ion (28) is applicable for nlany applications. 15 (28j SM=-10. log,,(.l) -0 In equaticn (28) o is a constant offset of about 8 to 10 .dB, typically about 3 dB, The SNR unit 1ZSb may output the SNR information to further sub-units of the signal 20 processing unit 120, for e:cample to a ilet~lodulatos unit for furtller equali.ziag the combined signal. According to another embodiment, the SNR information is used to adapt the step size parameter 1iP for the iteration algorithm or to generate a control -. signal to select a diversity selection scheme in a diversity combining unit. 25 According to a further embodiment, a window adjustment unit l2Sc measures a CBW (coherence bandwidth) of the current estimated channel t,r&nsfer functions and adjusts the observation window length L. for the iteration algoritb~n,~ vhicl-Ii.s. , lrly wxy of exa.mple, a MCFLMS or MCNFLMS (multi-channel nor~nalksdf requency least meen squares) algorithm, to keep the window Length as srsla.11 as possible. The s~indosv 30 adj~strnentu nit 125c may estimate the CBW (coherence bandwidth) by convclutio~~ of each current estimated channel transfer functions. The window : i f i j ~ ~ ~ l r n eu. ~nlitt. 125c may increase the window length L when tho messarrd CBW appears to he limited by the observation window. The windo\v adjustmefit 'unit 1 2 5 m~ a y decrease . . IJ~-!~~-Ij-L11I1I b. .- ! ..L(A I nn FAX N . : t4c '39 48YClIl3':: S, w"52 * 1n1O k r ~ v ~ ~ t Cdh< t v - 23 - SONI Corporation 43612 17.Q2.2012 . ---. . . -.-- . the window length L when the measured CBW appears to be ilotablg larger than the obscrvatian window theoretically allows. Adjusting the obeervation window llelagth mitigates both potential channel order uvercstinlatioil and Ioas of oxthogonaJity in case of fast-fading channels. 5 According to a further embodiment, the window adjustment unit 12 5c measures a CT (channel coherence time) on the basis of the previous tstinlatzd channtl transEer functions and adjusts the observation mi~ldnw length L accordingly. The window adjustment unit 125c may decrease the window length L when the rneasxlred CT 10 appears to be notably shorter than the observation window. The window adj~lstlllent 'unit 125c may increase the tvifidow length L when- the measured Cf appears to be notably larger than the observation window. Where the CBW and t.he CT estimation results lead to contradictory results for the observation window length i, the win do.^^ adjustment unit 125c adjusts the observation window length L such that the overall 15 system performance is optimized, e.g. a BER (bit error rate) is minimized. The signal processing unit 120 may uae the results for the estimated channel order (observation window length) for adjusting the computati.onaJ, complexity, for example by frequexy-domain sub-sampiing where the process is performed only every atii 20 Frequency tap with p 1 or by time-domain sub-ssmpling, where the pracess i s perfonxed only every qth signal block with observation window length L and q 1, wherein p and/or q are selected proportional to the e.stirnated channel order, by way of example. I 25 The signal processing unit 100 of the above-describe.d elnbodinlcnts and eo.ch of its sub-units may be realized in hudware, in s o f t ~ a r ao r ss s c.umt~iilation thei-itof. Some or all of the sub-units of the signal proctssing unit 100 may br, iateg-atcd in a common package, for example in an IC (integrated circuit), an ASIC (application specific circuit) or a DSP (digital signal processor). 30 Figure 5 refers to a method of operating a receiver apparatus. CIII the basie af previous estimated frequency-domain chan~ltl vectors, sparseness infornlation is obtained that is descriptive for a change of sparseness in the estin~aterl frequencySOW Corporation 63612 domain channel vectors (502). From the sparseness i.~ifo;r;mat.ioann d a. plur:a,lit.y of receive signals, estimates for frequency-domain channel vec.t0re of transmission channels are updated on the basis of 8. b],ind channel estimation dgorithrn exploiting both the sparseness information and cross-relations between pah-s of receive signals 5 (504). Thereby each receive signal is assigned to another one of the tran,srnissjzos channels and all receive signals originate from the same transmit signal. A further embodiment refers to a signal processing unit that includes an estimator unit, a cletecting unit and a power limiting unit. The estimator unit estimates, from a 10 plurality of receive signals, frequency-domain channel vectors of transmission channels OI-I the basis of an iter~tiveb lind cbennel csrinlation algorithm t>:ploiting cross-relations between pairs of receive signals, wherein each receive signal is assigned to another one of the transmission channels and 0riginat.e~f rom tthe same transmit signal. The detecting unit detects, on the basis of the fi-cqueni;y-domain 1.5 channel vcctors, a misconvergencp, behaviol-zr of the blind channcl estiinatiun . algorithm or an excess condition for out-of-bald energy i11 a frequer~cy range between an upper frequency limit of the receive signal a ~ i da sa~rplef requency and outputs s, control signal indicating a miscorivergence behaviuur cr an excess condition. The power limiting unit limits a signal energy in a frequdllcy band edge 20 area of the frequency-domain channel vectors it1 response to the coritrol sighal. The power limiting unit may limit the sigaal energy in the firequency band tclgc area adaptively in response to currently and/or prer~io~~esslyti mated frequency-domair; channel vectors, wherein the blind channel cstirnation algorithm proceecls with. the energy-limited estimated frequency-domain channel vectors. The blind challnel 25 estimation algorithm is improved, whether or not the blind channel estimation algorithm uses sparseness information c0ncernin.g a change of sparseness in the estirnatcd channel. A f ~ ~ r t hecmr bodiment refers to R, signal processing unit that includes an estimatol- 30 unit and an SNR unit. The estimator unit estimates from a plurality of receive signals, frequency-domaix channel vectors of transmission channels on the ba,sis of an iterative blind chaankl estimation algorithm exploiting cross-rela.tio1-i~ between pairs of receive signals, wherein each receive signal is assigned to another one of the transmission channels and originates from the sa,ine transmit signal. The SNR unit. outputs SNR information descriptive far current signal-to-noise ratio values of the channels on the basis of the values of a cost function minimized by the iterative blind channel estimation algorithm. The SNR unit may adjust a predefined step size 5 parameter of the blind channel estimation algorithm on the basis of the values of t.he cost func.tion. The blind channel estimation algorithm is improved, whether or not the blind channel estimation algorithm uses sparseness information. Each of the above embodiments improves existing blind channel estimation 10 algorithms. In combination with diversity combining, where s'tate-of-the-art receiver diversity techniques are bssed On partial Imowltdgc of the channel transfer function and therefore require a demodulator to be instantiated for each receive signc~l, the described embohiments rnay solve this ~hort-coming by means of ree.lizing hluid diversity combining before demodulation. The "blind" diversity co~llbinilel~xp loits the 15 fact that Ule knowledge of cross-relations between the channels suffices to apply diversity combining at a profit. The approach can be applied to a wide range of cammunicatlon systems. Costs for diversity receivers c3.n be significantly reduced. 112-i+bk-llllj-Ulb .ti: Zj Muller Hoitrnann FAX N:. :+4! 89 1391:1033 S, 037/0j2 / - * ~ H O F F M A N 8N: P ARTNER SONY Corpuration 63612 Claims 1. A signal processing unit comprising an estimator unit (120) configured to estimate, from a plurality of receive 5 signals, frequency-domain channel vectors of transmission channels on the basis of an ite,rative blind channel estimation algorithnl, which exploits cross-relations between pairs of receive signals and which is modified by a tern1 c.onsidering a change in a sparseness rneavure obtained from the eatime.terrl freque.ncy-dam air^ channel vectors, wherein each receive signal is assigned. to another one of the 10 tratlsnaissioa channels and originates from the same transmit signal. 2. The signal processing unit of claim 1, farther comprising a combining unit (130) corrfigured to combine the receive signals 011 the basis of combining coefficients derived from the estimated frequency-domain channel 15 vectors to obtain a combined receive signal repl-esenting an estimation of the transmit signal transmitted through the transmission chanflcls. 3. The signal processing unit of claim 2, whe.re.in the combining unit (130) is configured to apply ma-cimun ratio col-illiillhg b j ~ 20 determining, from the estimated frequency-domain channel vectors tir time-donldn cliannel impulse response: vectors derived therefrom, filter cotfficients for matched filters, tach matched filler matching with the estimated ckanacl itl~pulaer es~mnseo f one of the transmission channels. 25 4. The signal processing unit of any of the preceding cllair+s, wherein the estimator unit (120) is configured to estimate the frequency-domain channel vectors on the basis of an iterative algorithm nlinir~~izinag cost fu-nctian, wherein the iterative algorithm provides updating the current frequency-domain channe.1 vectors on tfie basis of previous frequency-domain channel vectors: a 30 predefined step size parameter, a current block error sequer-ce vector and a ci~rrent spars errtvs measure indicating a change in sparseness in the. cstina2tecl channel vectors. " a ?RFfi%ER SONY Corporation 636i2 5. The signal processing unit of cl~irn4 , wherein the block error sequence vector represents the Fourier-tro-nsfornled of block error signals between pairs 'of receive signale, each error signal re.s~.lltin:g frnlil e. deviation betweeh a result of 8 coav6lution of a receive signal received through a first 5 transmission channel with an estimated impulse response of a s e c o ~ ~trdah srnission -channel and a result of a convolution of a receive signal received through the second transmission channel with an estimated irn@ulse response of the first trensmiasioa channel, the iterative algorithm using se.cond-order c:ross-correlation iilformatioa about psirc; of the receive signals. 10 6. The sigild processing unit of m y o f claims 1 to 5, further conlprising a spwseness evaluation unit (127) configured to obtain the sparseness measure containing sparseness information for the estimated frequency-domain channel vectors, and 15 an adjustment unit (125s.) configured to adjust the estimated frf'tecpencydomain channel vectors on the basis of an adjustment term conr.aining the cbtained sparseness measvre. 7. The signal processing unit of any of claims 4 to 6, wherein 20 the iterative algorithm adjxst.8 tkc, updated estimated frcquer~oy-doriain cnannel vectors to increase the sparseness measure. 8. The signal processing unit of any of claims 4 to 7, wherei~; the iterative dgorithm adjusts the updated estimated freq~ieriv-drjrnkin 25 channel vectors. on the bash of an averaged time derivative of the sparseness measure. 9, The signal, processing unit uE any uf claims 4 ti, 8, wherei:~ the iterative algorithm adjusts the updated estilnatecl frequency-domain. 3C channd vectors such that for tach channel vector the sum of t.he cbanllel casfficicats is minimized and a signal energy for each channel vector is maintailled- 10. The signal processing unit of any of'claims 4 to 9, further comprising M w - 'I t0FFiviARbi rsc rm 1, NER SONY Corporation 63612 ax1 SNR unit (125b) configured to output SNR information descriptive for current signal-to-noise ratio values of the channels on the. basis of the values of the Cost functibn. 5 11. The signal processing unit of claim 10, wherein the SNR unit (125b) is c a n f i ~ r t dto adjust the predefined step size parameter on the basis of the values of the cost function. 12. The signal processing unit of any of claims 1 to 11, further comprising 10 a detecting unit (128a) configured to detect, on the basis of the ireq~enc;jrdomain channel vectors, a miuconvergence behaviour of the blind channel estimation - alg.o rithm or an excess condition for out-of-band energy in a frequency range between an upper frequency limit of the receive signal and a sample frequency and to output a control signal indicating a misconvergcnce behaviour or ail excess 15 condition, 13. The signal processing unit of claim 12, further comprising -. a power limiting unit (128b) configured to lirnit a signal energy in frequency band edge areas of the estimated frequency:dotnain channel vectors i l l response to 20 the control signal, wherein the iterative blind channel estimation algc~sithin proceeds with the energy-limited estimated frequency-domain channel vectors. 14. The signal processing unit of any of the preceding claims, f ~ ~ r t hceorm prising a window adjustment unit (12.5~) coafigurtd to measure a ccsb.erence 25 bandwidth of the cutrent estimated channel vectors md to adjust an ohservatiori window length for the blind channel estimation algoritk~m on the Isasis of the measured coherence bandwidths, 15. The signal prpcesshg unit of claim 14, wherein 3 0 The window adjustment unit (125~)is configured to increase the window length L when the rneasvred coherence bandwidth is limited by the ubset-vatinn window and to decrease the window length when the measured coherence bandwidth is notably larger than the observa,tion wi1lcl.o~ allows. 16. The signal processing unit i f any of claims 14 ta 15, whereih the window adjustment unit (125c) is configured to measure a channel coherence time on the basis of the previous estimated channel vectors and to adjust 5 the abservatidn window length on the basis of the measured chahne.1 coherence time. 17, The signal processing unit of claim 16, wherein the window adjustment unit (12Sc) is configured to decrea.se the window length when the measured CT appears to be ootably shorter than the ohaervatiori 10 window and to increase the window length L when the measured CT appears to be notably larger than the observation window. 18. AQ integrated circuit comprising the signal processing unit accordilig to any of the preceding claims- 15 19. An tlectro~licd evice comprising tne signal processing unit according to any of claims 1 to 17; and a plurality of tuner circuits (20$), each tuner circuit (2113) c.onfigured to tune to a carrier frequency of a transmission signal and to output al-L analogue receive 23 signal, wherein ezch analogue receive signal is assigned to one of the transmission channels. 20. A ixethod of operating a receiver apparatus, the method comprising obtaining, on the basis of previous est.imated frequency-doniain c h a ~ i e l 25 vectors, a sparseness measure containing informat.j.on desc.rlptive for a sparsel-less of the estimated frequency-domain channel vectors; end estimating, using an estimator unit, from a plurality of receive signals curent' frequency-domain channel vectors of transmission channels on the basis af an iterative blind channel estimation algoritfim exploiting c.ross-relations bePweeil pairs 30 of receive signals, wherein the iterative blind channel estimation algcsrithrn is r~odified on the basis of s change of the sparseness measure obtainecl fro~ri th-r; estimated frequency-domain channel vectors, and wherein each. rec.eive sigr~al is MJlUAX - i-iCFFKPJW & PARTNER - .>.> ('3 - 80KY Corporatian 63612 17 0'5.2015 .---- assigned to another one of the transmission channels and originates from the same trtmsmit signal. 31. The method according to claim 20, ftlrtller comprising 5 combining, using a combining unit, the receive signals an the basis of: combining cocfricie.nts derived from rhe estimated frequency-domain channr.1 ilectors to obtain a combined receive signal representing an cstirnatinn of the traneuiit signd transmitted through the transmission channels. . , 10 22. The method according to claim 21, wherein combining incl.udes determining, from the estimated Frequency-domain channel ve12tnrs or timcdomain channel impulse response vectors derived therefrom, filter coefficients for matched filters, each matched filter matching with the estimated channel iml~dlse response of one of the transrnissiop chapnels; and 15 applying a maximum ratio carnbining scheme on the receive signals. 23. The method according to any of claims 20 to 22, wherein estimating the frequency-domain channel vectors includes applying an iterative algorithm minimizing a cost function, wherein the iterative algorithm 20 provides updating the freqxency-domain channel vectors on the basis {if a current frequency-doma* channel vector, a predefined step eize parameter, a currenr block error sequence vector and s current sparseness 1rIeaslu.e. 24. The method according to claim 23, wherein 2 5 the black error sequence vector represents the Fourier-tra~~sforrncodf block error sigzals between, pairs of receive signals, each error sig11.d ~.t'~Ultiffrgo m 5. deviation 'between a result of R c.on~olu~ioofn a recei.\re signal received thmugh a first translzlission channel with an estimated Impulse response of a second tranami:3sion channel and a result of s convolution of a receive signal received thrwgh the secolld 3U transmission channel with an estimated impulse response of the first transmission channel; and the iterative algorithm -ases scco~&order stochast;~ cross-correlation inforlnaticn about pairs of the receive signals. MULL'I- a PARTNER SONY Carporation 25. The method according to any of claims 20 to 24, wherein - the sparseness measure is obtained using a sparseness evaluation (127j evaluating the sparseness measure on the basis of the e.stirriated frequency-domain 5 channel vectors, and the estimated frequency-domain channel vectors are adjusted, before a next iteration, on the basis of an adjustment term containing the obtained sparseness measure. 10 26. A signal processing .unit comprising an estimator unit (120) configured to estimate, from a pl~~ralitoyf receive signals, frequency-domain channel vectors of transmission channels on the basis of an iterative blind c h a n ~ e l estimation algorithm esploiting crass-relations between pairs of receive signals, wherein each receive signal is assigned to anotl-ler one of the 15 transanission channels and originates from the same transmit signal, a detecting unit (128~~co) nfigured to detect, on the basis of the frec4uencydomain channel vectors a 11zisconverge11,ce kehavioiu- of the bli1-d channel estimation algorithm or an excess condition for out-of-band energy in a frequency range between an upper frequcney limit of the receive s'igaal snd a sample fse~quency and to 20 output a control signal indicating a rnisconvergenc:e behnviour or nn excess condition, aad. a power limiting cnit (128b) configured to limit a signal energy in frequency band edge areas of the frequency-domain channel vectors in response to the co~ltrol signal, wherein the iterative blind channel estimation algorithm proceeds xii-ith the 25 energy-limited estimated frequency-domain channel vectors.. 27. A signal procesaing unit comprising an estimator unit (120) configured to p,st.ima,t,e, from a plurality of receive signals, frequency-domain channel vectors of transmissioil channels on the basis OS 30 a blind clxunnel estimatioi~ algorithm exploiting c.ross-relatioils bctweefl peirs of receive signals, wherein each receive signal is assigned to another one of the trmnsnlission channels and originates from the same tranvmit signal, and UJ I k J - L U I J UID - U . L r r n 1~4 . . . I q; UJ Y O J I , I U J ) #J, IJY.:I.~ UIL M w kkJf-FMANN & PARTNER * - 87 ,-, -30 - SONY Corporation 63612 17.09 .poi3 .-._ an SNR unit 125h configured t o output SNR information descriptive -for current signal-to-noiae ratio values of the channels on the basis of the vLducs of a cost function nlinimized by the blind channel estimation algorithm.