Blind Deconvolution of Multi-Input Single-Output Systems Using the Distribution of Point Distances

Size: px
Start display at page:

Download "Blind Deconvolution of Multi-Input Single-Output Systems Using the Distribution of Point Distances"

Transcription

1 J Sign Process Syst (2) 65: DOI.7/s Blind Deconvolution of Multi-Input Single-Output Systems Using the Distribution of Point Distances Konstantinos Diamantaras Theophilos Papadimitriou Received: 22 January 2 / Revised: 27 October 2 / Accepted: 27 October 2 / Published online: 3 November 2 Springer Science+Business Media, LLC 2 Abstract In this paper we present a new blind identification and source separation method for linear Multi Input Single Output (MISO) convolutive systems driven by PAM sources. Our method is based on the estimation of the output difference distribution for pairs of outputs. We show that the most likely differences (not counting the zero difference) are the ones corresponding to the columns of the mixing matrix (upto a sign). The columns can be arranged in the correct order by using the block-toeplitz property of the transfer matrix. Thus the problem is transformed into the density estimation problem. The method is conceptually simple and can work with relatively small data sets although it is exponentially complex with the channel length or the number of input signals. Keywords Blind deconvolution Finite alphabet sources Combinatorial methods Introduction Blind Source Separation (BSS) refers to the recovery of n unknown signals, using m recorded mixtures. K. Diamantaras (B) Department of Informatics, TEI of Thessaloniki, Sindos, 574, Greece kdiamant@it.teithe.gr Th. Papadimitriou Department of International Economic Relations & Development, Democritus University of Thrace, Komotini, 69, Greece papadimi@ierd.duth.gr In BSS we try to develop methods needing the less possible aprioriknowledge on the system; the systems are never totally blind. In linear BSS, the mixing can be memoryless or convolutive and each case is treated quite differently. The mixing operator in memoryless systems is a constant matrix and there is no time shift of the source signals. However the instantaneous model cannot describe correctly the mixing procedure in an echoic environment where the same source arrives in the sensors with various delays. These cases are treated using the Multi-Input Multi-Output (MIMO) convolutive model. So the researcher must know if the mixture signals he is working with belong to the former (treated with instantaneous BSS methods) or the latter group (treated with multichannel blind deconvolution methods). Fortunately, the mixing procedure can be assumed according to the nature of the system and the desired accuracy of the recovered signals. For example, we do know that in recording studios the mixtures taken by a set of microphones are convolutive. However if the microphones are close to the sound sources, we can assume that wall distortions and echo are negligible and use an instantaneous BSS method yielding lower though acceptable accuracy. The instantaneous problem has been extensively studied in the past using Second-Order Statistics (AMUSE [, SOBI [2, OPCA [3) and Higher-Order Statistics (JADE [4, ICA [5). The convolutive problem is much more complex and is still an active research topic. The first methods aimed at disintegrating the multichannel system into a set of simpler SIMO (Single- Input Multi-Output) systems, which are easier to solve [6, 7. The special case of i.i.d. has been studied in [8, by exploiting the structure of cumulants. Colored sources

2 526 J Sign Process Syst (2) 65: have been treated in the past using both second order [9, and higher order statistics [. There are some very useful properties in the output sequence when the system is driven by finite alphabet sources. Typically such methods exploit the combinatorial analysis and the geometric properties of the observations constellation. Diamantaras e.a. studied the memoryless MISO (Multi-Input Single-Output) system driven by binary antipodal sources in [2. The system filter is recursively deflated, estimating eventually the filter and the source signals. We extended the basic idea to the convolutive case in [3. Richer finite alphabet sources, i.e. Pulse Amplitude Modulated (PAM) and Quadrature Amplitude Modulated (QAM) sources were investigated in [4. The proposed method exploits the distribution of the distances between the observations cluster centers. The mixing vectors can be estimated through a combinatorial analysis of the distances histogram. When the mixtures are more than the sources the system is called overdetermined (m > n); the opposite case system where the sensors are less than the sources is underdetermined (n > m). In the BSS framework there are several established and well-known methods treating systems of the former case, and much fewer treating the more demanding problem described in the latter case. One of the most difficult BSS problems is formed when many sources are mixed through convolution in just one mixture signal. The problem is described by the Multi-Input Single-Output model (MISO) and is unsolved in its general form. We present in this paper a new method for the blind deconvolution of PAM driven MISO systems. The basic idea relates to the probability distribution of the output differences. The second most probable differences are related to the columns of the system transfer matrix and they can be used so we can estimate the mixing filters. The paper is organized as follows: Section 2 describes the problem formulation, Section 3 gives the basic idea and the proposed method including transfer matrix estimation, source recovery and dealing with noise. Section 4 presents simulation examples for noiseless and noisy cases, while Section 5 discusses computational complexity issues. We finally conclude in Section 6. 2 Problem Formulation A Multi-Input Single-Output (MISO) convolutive system with n discrete inputs s i (k), i =,, n, and equal number of L-tap real filters h i (l), i =,, n, l =,, L, is described by Eq. x(k) = n L h i (l)s i (k l). () i= l= The input signals s i, the number of sources n, and the filters h i are unknown, but for the sake of simplicity let us assume that the filter length, L, is given. The estimation of L is treated in Section 3.. Eq. can be rewritten as L x(k) = h(l) T s(k l) (2) where l= h(l) =[h (l),, h n (l) T, l =,,, L, s(k) =[s (k),, s n (k) T. Suppose that the inputs are Pulse Amplitude Modulated (PAM) with M levels so s i (k) {v + mt m =,, M, v IR}. Without loss of generality, assume that the gap between successive levels is T =, so the discrete input symbol set is V s ={v m = v + m m =,, M } (3) Let us define the range of x(k) as follows V x ={x(k) from Eq. s i V s }. (4) Obviously, the cardinality of V x is M nl provided that there are no repeated values in V x. Taking a timewindow of length W starting at time k and stacking the previous outputs into a vector x(k) we obtain x(k) =[x(k),, x(k W + ) T = Hs(k), (5) where h() T h(l ) T h() T h(l ) T H=......, h() T h(l ) T s(k) =[ s(k) T, s(k ) T,, s(k W L + 2) T T. (6) (7) The dimensions of matrix H and vector s are (W Q) and (Q ), respectively, where Q = n(w + L ). (8)

3 J Sign Process Syst (2) 65: As before, we can define the sets of values for s and x as follows V s ={s(k) from Eq. 7 s i V s }=V Q s (9) V x ={x(k) from Eq. 5 s V s }. () The cardinalities are V s = V x =M Q, assuming that there are no repeated output values. We assume that the sources are independent white sequences, i.e., they satisfy the following conditions: A. The signals s i are independent to each other A.2 The samples s i (k) are i.i.d. We also assume that the symbols are emitted with equal probabilities, i.e. A.3 The symbol transmission probabilities are uniform: p s = Prob(s i (k) = v m ) = /M, k, i, m. () Assumption [A.3 is instrumental in the subsequent development of our method which is based on the probability distribution of the output differences. This probability distribution is shown to have interesting properties revealing the structure of the mixing system, provided that the source symbol distribution if uniform. 3 Output Difference Distribution Let us consider the difference between any two output vectors x(k), x(k ), at different time instances k = k, d(k, k ) = x(k) x(k ) = Hu(k, k ), (2) u(k, k ) = s(k) s(k ). (3) The distribution of d obviously depends on the distribution of u and u(k, k ) =[ū(k, k ) T,, ū(k W L + 2, where k W L + 2) T T (4) ū(k, k ) = s(k) s(k ) =[δ (k, k ),,δ n (k, k ) T (5) with δ i being the input differences δ i (k, k ) = s i (k) s i (k ). (6) Clearly, the range of values for δ i is the set U ={ (M ), (M 2),,(M 2), (M )}. (7) Then from Eq. 5 we have u U Q. In the sequel, we shall drop the indexes k, k, due to input stationarity. All δ i have the same distribution, call it p δ (), forall i, but this distribution is no longer uniform. Since the input symbols are transmitted with equal probability p s = /M, p δ is triangular. Indeed, we have: p δ (β) = Prob(δ i = β) = (M β )/M 2, β U. (8) So, the most probable difference is δ i =, with probability p δ () = M/M 2 = /M. The second most probable difference is δ i =±, with probability p δ (±) = (M )/M 2, the third most probable difference is δ i =±2, with probability p δ (±2) = (M 2)/M 2, andsoon. Based on the above, we can compute the distribution of the vector u given that the elements of u are statistically independent p u ([u,, u Q T ) = Prob(u =[u,, u Q T ) = Q i= p δ(u i ) (9) We make the following assumption relating the input and the output difference vectors, A.4 For all u, u U Q,ifu = u then Hu = Hu. If a matrix H satisfies [A.4 then we shall say that its columns h i are Delta-independent, i.e. they satisfy the following property: Q α i h i =, α i α i =, i (2) i= where ={δ δ 2 δ,δ 2 U} ={ 2(M ),, 2(M )}. (2) Property 2 is like linear independence except that the coefficients α i are constrained to fall into the discrete set. If the elements of H are random reals then

4 528 J Sign Process Syst (2) 65: Delta-independence holds with probability and so assumption [A.4 is satisfied. The most important consequence of [A.4 is that the probability of u propagates to the probability of d in the following sense Prob(d = Hu) = Prob(u). (22) Based on Eq. 9 and due to the triangular distribution of δ i the most likely vector u is u =[,, T = with probability p u () = /M Q.AccordingtoEq.22 the corresponding most likely output vector is d = and Prob(d = ) = /M Q. (23) The second most likely set of input vectors are those belonging to the set U ={u =[,,, u i,,, T u i =±, for some i : i Q} (24) All vectors in U have the same probability which, basedoneq.9,is p u (u U ) =[/M Q (M )/M 2 = (M )/M Q+. (25) There are 2Q such vectors. The output vectors d corresponding to those u U form the set D ={d = Hu u U } (26) and can be expressed as d =±h i, i =, 2,, Q (27) where h i is the i-th column of H. AccordingtoEq.22 these output differences have the same probability as the expression in Eq. 25: p d (d D ) =[/M Q (M )/M 2 = (M )/M Q+. (28) It is not difficult to see that all other input differences (except for u = and u U ) have smaller probabilities. Therefore, the corresponding output differences also have smaller probabilities. If we have the output difference distribution p d then simple sorting of the probabilities will reveal the columns of the mixing matrix H, upto a sign and without the correct order. In summary, the most likely output difference is d = with probability /M Q, the second 2Q most likely output differences are the vectors d =±h i,fori =, 2,, Q, each with probability (M )/M Q+. 3. Estimating Filter Length The filter length L can be readily estimated using the output covariance function r x (l) = E{x(k)x(k l)} which is obviously zero for l > L. Since r(l ) = h(l ) T h, if the product h(l ) T h is non-zero then L can be estimated by finding the smallest l > for which r x (l) =. 3.2 Recovering Column Order From the analysis in the previous sections it follows that the second 2Q most likely output differences d i, i =,, 2Q, form the set {h,, h Q, h,, h Q }. According to Eq. 6 the matrix H has a block-toepitz structure. Therefore, the first L columns forming the W L submatrix h() T. are filled with zeros everywhere from row 2 to row W. Similarly, the last L rows are filled with zeros everywhere from row to row W. Since we have estimates of the columns, even though without proper ordering, we can identify those vectors d i which are filled with zeros everywhere from row 2 to W. They should correspond to the first L columns of H and their negatives, so there should exist 2L such vectors. So far we know that the vectors d i estimate the columns of H (and their negatives) but without any prescribed order. If we want to estimate H we need to recover this ordering information. To that end, we exploit again the block-toeplitz structure of H, which is expressed by the following property h i, j = h i+q, j+lq for all q (29) We start with the vectors with zeros everywhere from row 2 to row W since we know that they belong to the first L columns. Then, following Eq. 29 we arrange the vectors d i so that d i, j = d i+q, j+lq,forallq. This will recover the mixing vectors h(l) upto a permutation π of their elements and a sign σ i, ie. we shall obtain ĥ(l) =[σ h π() (l), σ 2 h π(2) (l),,σ n h π(n) (l) T, for all l =,, L. (3)

5 J Sign Process Syst (2) 65: where σ i =±. We can tolerate this permutation and sign issue however, since it corresponds to a permutation and sign of the source signals, and it is well known that the order and the sign of the source signals cannot be recovered in a blind source separation scenario. The following example clarifies how H is estimated by arranging d i as described above. Example Consider the following system (n = 2, L = 2, W = 2,soQ = 6): [ x(k) = s(k) Assume binary inputs (M = 2), then the difference vector d = is the most frequent one with probability /M Q =.56 the next 2Q = 2 most frequent difference vectors d, each with probability (M )/M Q+ =.78, are the following (in no particular order) d = d 4 = d 7 = d = [.66 [.2.43 [ [.43 d 2 = d 5 = d 8 = [.66 d = [.29 [ [.2 d 3 = [.2.43 [ d 6 =.29 d 9 = [.43 [ d 2 =.2 Start with vector d which has a zero in row 2: this is column. We then arrange d 8 and d 6 in columns + L = 3 and + 2L = 5 since they satisfy the Toeplitz property [d =[d 8 2 and [d 8 =[d 6 2, respectively. So far, our estimate of the mixing matrix H is [ We repeat the procedure in an entirely analogous way starting with vector d 9 which has a zero in row 2: this is now column 2. (Vector d 2 is not eligible, although it also has a zero in row 2, because d 2 = d ). We then arrange d 4 and d 2 in columns 2 + L = 4 and 2 + 2L = 6 since they satisfy the Toeplitz property [d =[d 4 2 and [d 4 =[d 2 2, respectively. Our final estimate of the mixing matrix H is Ĥ =. [ Now we can form the estimates of the filter taps (mixing vectors) h(l) from the first row of Ĥ, namely, ĥ() =[ĥ,, ĥ,2 T =[.66,.43 T ĥ() =[ĥ,3, ĥ,4 T =[.29,.2 T. Notice that the order of the parameters is swapped compared to the true filter taps: h() =[.43,.66 T h() =[.2,.29 T. The parameters corresponding to the second source (.66,.29) are put first in our estimate while the parameters corresponding to the first source (.43,.2) are put second. At the same time the sign of the parameters corresponding to the first source is flipped. This simply means that when we shall extract the sources, the first source will be extracted second and with a flipped sign, while the second source will be extracted first. This is the typical sign and order ambiguity encountered in all blind source separation problems. The preceding analysis leads to the following algorithm for blind MISO channel identification: Algorithm Blind MISO channel identification Compute the output differences d(k, k ) for all pairs of indexes k, k Estimate the distribution p d of d Sort the probabilities ˆp d in decreasing order. Let d i, i =,, 2Q, be the values of d with probabilities ranging from the 2nd highest to the (2Q + )-th highest probability. Arrange the vectors d i into a block-toeplitz matrix as described above by forcing the satisfaction of property 29. From the elements of the submatrices we obtain our estimates of the mixing filters. 3.3 Recovering the Sources Given the estimates of the filter taps h(l) we can estimate the sources by minimizing [ŝ(k),, ŝ(k L + ) =arg min s,, s L V n L x(k) ĥ(l) s l (3) l=

6 53 J Sign Process Syst (2) 65: Since the input alphabet, V, is discrete the simplest way of minimizing Eq. 3 is by exhaustive search in the input space. Multiple estimates ŝ(k) of s(k) will be obtained by performing the above minimization for x(k), x(k + ),..., x(k + L ). These estimates are averaged out to yield the final estimate of s(k). 3.4 The Addition of Noise In the presence of additive noise e(k), the model becomes x(k) = Hs(k) + e(k) (32) where e(k) =[e(k),, e(k m + ) T. Similarly, the difference model also contains an additive noise term d(k, k ) = Hu(k, k ) + d e (k, k ) (33) where d e (k, k ) = e(k) e(k ). Using the Bayes rule we have p(d) = u = u p(d u)p(u) p de (d Hu)p(u) (34) So the distribution of d is a mixture of the distributions p de centered at the points Hu andscaledbythefactors p(u). If, for example, the noise is zero-mean, Gaussian, then p(d) is a Gaussian mixture. Our preceding analysis is still valid regarding p(u). Therefore, the tallest Gaussian bell is the one associated with u =, while the next 2Q tallest bells are the ones associated with u U. If we have an accurate estimate of the bell centers then we can apply Algorithm directly and obtain an estimate of the mixing matrix. The estimation of the distribution p(d) using a parametric method, such as the EM algorithm [5, has the advantage that it can produce good estimates of the centers even if the Gaussian bells merge with each other. The disadvantage is the fact that there can be, potentially, a very large number of bells (M Q ) while we are really interested in only (2Q) of them. The alternative is the use of nonparametric methods such as Parzen window smoothing [6, chapter 4. This approach is computationally more efficient although it can be quite demanding in memory especially for m > 4. In this work this will be our method of choice. 4Results We conducted various sets of experiments following noiseless and noisy MISO scenarios. We will present results on both cases..6 x p(d) Figure (Left) The 3 highest estimated probabilities for the vectors d arranged in decreasing order. The probabilities ranking from 2 to 7 (dark bullets) correspond to the vectors estimating the columns h(i) of H. (Right) The vectors d corresponding to the 3 highest probabilities. As in the left plot, the dark bullets correspond to the vectors d with probabilities ranking from 2 to 7. The crosses + indicate the position of the true vectors ±h(i).

7 J Sign Process Syst (2) 65: Noiseless Scenario 4.. Case The first noiseless experiment featured n = 2 independent sources with M = 3 PAM levels. The L = 3 filtertaps are: [.4792 h() =.6623 [.2838 h(2) =.923, h() = [.749,.3657 The value W = 2 was used for the size of the output vector and the probability distribution p d () was estimated using a simple frequency histogram on the output vector differences based on N = 2,5 data points. We have Q = 2(L + ) = 8 so we are looking for the difference vectors with frequencies from the 2nd highest to the (2Q + )-th = 7-th highest. After rearranging the vectors as described above we obtain the following filter estimation: ĥ() = [.48.66, ĥ() = [.7.37 [.28, ĥ(2) =.9 In Fig. (right) we show the vectors with the 3 highest estimated probabilities and their probability values (left). There is an obvious gap between the st and 2nd probabilities and also between the 7th and the 8th marking the beginning and the end of the probability values we are interested in. Ideally, the highest probability should be equal to /M Q = , and the probabilities from 2 to 7 should be equal to (M )/M Q+ =.6 4.Wecanseethatthe true and estimated clusters (the black dots and the crosses in Fig., right) practically coincide yielding very good estimates of the vectors ±h(i). The source reconstruction according to (3) gives Symbol Error Rate: SER = Case 2 In the second noiseless experiment we used n = 3 independent sources with M = 2 PAM levels. The filter taps (L = 4) were: h() =.3444, h() =.7, h(2) = , h(3) = The output vector had W = 2 size and following the presented method we estimated the probability distribution p d () using a simple frequency histogram on the output vector differences based on N = 6, data points. We have Q = 3(L + ) = 5 so we are looking for the difference vectors with frequencies from the 2nd highest to the (2Q + )-th = 3-st highest. 4 x p(d) Figure 2 (Left) The 5 highest estimated probabilities for the vectors d arranged in decreasing order. The probabilities ranking from 2 to 3 (dark bullets) correspond to the vectors estimating the columns h(i) of H. (Right) The vectors d corresponding to the 5 highest probabilities. As in the left plot, the dark bullets correspond to the vectors d with probabilities ranking from 2 to 3. The crosses + indicate the position of the true vectors ±h(i).

8 532 J Sign Process Syst (2) 65: Figure 3 (Left) Density estimation using kernel smoothing (Right) Contour of the density function indicating the highest peaks (larger circles correspond to higher peaks). Using our method we obtained the following tap estimates:.59 ĥ() =.34,.95 ĥ() =.7, ĥ(2) =.87, ĥ(3) = leading to perfect signal reconstruction (SER = ). In Fig. 2 we show the vectors with the 5 highest estimated probabilities. Again there is an obvious gap between the st and 2nd probabilities and also between the 3st and the 32nd marking the beginning and the end of the probability values we are interested in. Again the cluster estimates (black dots in Fig. 2, right) and the true values (crosses) practically coincide thus obtaining perfect source reconstruction. 4.2 Noisy Case We then tested the method in a noisy scenario with n = 2 independent binary sources (M = 2) and output timewindow size W = 2. The Signal-to-noise ratio (SNR) was set at 3dB. The mixing filters of length L = 2 were chosen randomly and they are shown below: h() = [.332.2, h() = [ x -4.5 x p(d).8 p(d) Figure 4 The 3 highest estimated probabilities for the vectors d arranged in decreasing order for W = 2 (left) andw = 3 (right).

9 J Sign Process Syst (2) 65: The density was estimated using N =,5 data samples and their pairwise differences. The non-parametric smoothing method was used with Gaussian kernels of width σ =.. We then counted the highest peaks of the estimated density and using the proposed methods we obtained the following estimates ĥ() = [.33.2, ĥ() = [ Subsequently, the source reconstruction by Eq. 3 yields SER =.27. Figure 3 shows the estimated disribution and the locations of its highest peaks. 5 Discussion A. Computational Issues Since the probability of the vectors we are looking for is (M )/M Q+, it follows that we need at least O(M Q+ ) difference samples for a reliable estimation of the probability density. Luckily, a data set {x(k)} with N samples yields N(N ) = O(N 2 ) differences, so the size of the original data set can be only O(M (Q+)/2 ). Still, since Q = n(w + L ), we conclude that the method has exponential complexity with respect to W, n, andl. B. The Window Size W Clearly the choice of the parameter W affects the performance of the algorithm. The increase of W gives more dimensions to the observation vector x but this does not lead to performance increase. The reason is that the length of the input vector s is Q = n(w + L ) and it also increases linearly with W byafactorn. So, for example, in a problem with M = 2, n = 3, andl = 4, the size of s is Q = 5, forw = 2, while for W = 3 we have Q = 8. So, although the output space dimension has increased from 2 to 3, the input space dimension has increased from 5 to 8 and the cardinality of the output state set V x has increased from 2 5 to 2 8. Therefore, the output space has more clusters and more data samples are needed in order to achieve the same performance. Figure 4 shows the sorted probabilities for experiment Case in Section 4.. with W = 2 and W = 3 for N = 25 samples. We can see that the probabilities are not clearly separated for W = 3 (right), as opposed to the case W = 2 (left). C. The Ef fect of Noise The method proposed in this paper is quite sensitive to noise. The reason is that it depends heavily on the correct clustering of the output values and the accurate estimation of the difference probabilities. In the case of low SNR the clustering becomes difficult as clusters of noisy observations tend to merge together into bigger clouds. In a future work we shall consider applying post-processing methods to. improve performance by imposing specific constraints on the estimated vectors. However, this is beyond the scope of the present paper. 6 Conclusions We presented a new blind deconvolution method for real, linear Multi Input Single Output (MISO) convolutive systems when the sources are M-level PAM signals. Our method is based on the observation that the most likely input vector differences (except for zero) are the vectors having all zeros except for one non-zero element equal to ±. Thus by estimating the distribution of the differences of the output vectors we are able to obtain the columns of the system transfer matrix without correct order and upto a sign. Fortunately, the correct order can be recovered taking advantage of the Toeplitz structure of the matrix. Once the system is estimated the sources can be recovered by a straightforward minimization procedure. References. Tong, L., Liu, R., Soon, V. C., & Huang, Y. F. (99). Indeterminacy and identifiability of blind identification. IEEE Transactions on Circuits and Systems I, 38(5), Belouchrani, A., Abed-Meraim, K., Cardoso, J.-F., & Moulines, E. (997). A Blind Source Separation Technique Using Second-Order Statistics. IEEE Transactions on Signal Processing, 45(2), Diamantaras, K. I., & Papadimitriou, Th. (23). Blind signal separation using oriented PCA neural models. In Proc. IEEE Int. Conference on Acoustic, Speech and Signal Processing (ICASSP-3), Hong Kong. 4. Cardoso, J.-F., & Souloumiac, A. (993). Blind beamforming for non Gaussian signals. IEE Proceesings-F, 4(6), Hyvärinen, A. (999) Fast and robust fixed-point algorithms for independent component analysis. IEEE Transactions on Neural Networks, (3), Inouye, Y., & Hirano, K. (997). Cumulant-based blind identification of linear multi-input multi-output systems driven by colored systems. IEEE Transactions on Signal Processing, 45(6), Tugnait, J. K. (998). On blind separation of convolutive mixtures of independent linear signals in unknown additive noise. IEEE Transactions on Signal Processing, 46(), Chen, B., & Petropulu, A. P. (2). Frequency domain blind MIMO system identification based on second- and higherorder statistics. IEEE Transactions on Signal Processing, 49(8), Diamantaras, K. I., Petropulu, A. P., & Chen, B. (2). Blind two-input two-output FIR channel identification based on second-order statistics. IEEE Transactions on Signal Processing, 48(2), An, S., Hua, Y., Manton, J. H., & Fang, Z. (25). Group decorrelation enhanced subspace method for identifying FIR

10 534 J Sign Process Syst (2) 65: MIMO channels driven by unknown uncorrelated colored sources. IEEE Transactions on Signal Processing, 53(2), Shamsunder, S., & Giannakis, G. B. (997). Multichannel blind signal separation and reconstruction. IEEE Transactions on Speech and Audio Processing, 5, Diamantaras, K. I., & Chassioti, E. (2). Blind separation of n binary sources from one observation: A deterministic approach. In Proc. Second Int. Workshop on ICA and BSS (pp ). Helsinki, Finland. 3. Diamantaras, K., & Papadimitriou, T. (26). Blind deconvolution of multi-input single-output systems with binary sources. IEEE Transactions on Signal Processing, 54(), Diamantaras, K. I., & Papadimitriou, Th. (29). Histogram based blind identification and source separation from linear instantaneous mixtures. In Lecture Notes in Computer Science Proceedings of the 8th Int. Conference on Independent Component Analysis and Signal Separation (Vol. 544, pp ). Paraty, Brazil. 5. Bishop, C. M. (995). Neural netowrks for pattern recognition. Oxford University Press. 6. Duda, R. O., Hart, P. E., & Stork, D. G. (2). Pattern classif ication. Wiley, NY. editor for the IEEE Transactions on Signal Processing, IEEE Signal Processing Letters and the Springer Journal of Signal Processing Systems. In 997, he was a co-recipient of the IEEE Best Paper Award in the area of Neural Networks for Signal Processing. He is the author of the book Principal Component Neural Networks: Theory and Applications, co-authored with S.Y.Kung (New York: Wiley, 996). He is currently chairman of the IEEE Machine Learning for Signal Processing (MLSP) Technical Committee and from 25 to 28 he was a member of the IEEE Signal Processing Theory and Methods (SPTM) TC. He has been the chairman of the 27 IEEE Int. Workshop on Machine Learning for Signal Processing (MLSP27) and the 2th International Conference on Artificial Neural Networks (ICANN2) both held in Thessaloniki, Greece, and has also served as technical committee member for various international signal processing and neural networks conferences. He is a member of the Technical Chamber of Greece. Konstantinos Diamantaras was born in Athens, Greece. He received his Diploma in Electrical Engineering from the National Technical University of Athens in 987 and the Ph.D. degree, also in electrical engineering, from Princeton University, Princeton, NJ, in 992. Subsequently, he joined Siemens Corp. Research, Princeton, as a post-doctoral researcher, and in 995, he worked as a researcher with the Department of Electrical and Computer Engineering, Aristotle University of Thessaloniki, Thessaloniki, Greece. Since 998, he has been with the Department of Informatics, Technological Education Institute of Thessaloniki, where he currently holds the position of professor. His research interests include signal processing, neural networks, image processing, and parallel processing. Dr. Diamantaras is a Senior Member of IEEE and also serves as associate Theophilos Papadimitriou was born in Thessaloniki, Greece, in 972. He received the Diploma degree in mathematics from the Aristotle University of Thessaloniki, Greece, and the D.E.A. A.R.A.V.I.S (Automatique, Robotique, Algorithmique, Vision, Image, Signale) degree from the University of Nice-Sophia Antipolis, France, both in 996 and the Ph.D. degree in electrical engineering from the Aristotle University of Thessaloniki in 2. In 2, he joined the Department of International Economic Relations and Development, Democritos University of Thrace, Komotini, Greece, where, he served as a lecturer (22 28). Currently he holds the position of Assistant Professor in the same department. Dr. Papadimitriou served as a reviewer for various publications and as a member to scientific committees member for Conferences and workshops. In 27 he was a member of the organizing committee of the IEEE Workshop on Machine Learning for Signal Processing held in Thessaloniki, Greece. Theophilos Papadimitriou current research interests include digital signal and image processing, data analysis, and neural networks.

ORIENTED PCA AND BLIND SIGNAL SEPARATION

ORIENTED PCA AND BLIND SIGNAL SEPARATION ORIENTED PCA AND BLIND SIGNAL SEPARATION K. I. Diamantaras Department of Informatics TEI of Thessaloniki Sindos 54101, Greece kdiamant@it.teithe.gr Th. Papadimitriou Department of Int. Economic Relat.

More information

Subspace-based Channel Shortening for the Blind Separation of Convolutive Mixtures

Subspace-based Channel Shortening for the Blind Separation of Convolutive Mixtures Subspace-based Channel Shortening for the Blind Separation of Convolutive Mixtures Konstantinos I. Diamantaras, Member, IEEE and Theophilos Papadimitriou, Member, IEEE IEEE Transactions on Signal Processing,

More information

Blind Deconvolution of Multi-Input Single-Output Systems with Binary Sources

Blind Deconvolution of Multi-Input Single-Output Systems with Binary Sources Blind Deconvolution of Multi-Input Single-Output Systems with Binary Sources Konstantinos I. Diamantaras, Member, IEEE and Theophilos Papadimitriou, Member, IEEE IEEE Transactions on Signal Processing,

More information

Determining the Optimal Decision Delay Parameter for a Linear Equalizer

Determining the Optimal Decision Delay Parameter for a Linear Equalizer International Journal of Automation and Computing 1 (2005) 20-24 Determining the Optimal Decision Delay Parameter for a Linear Equalizer Eng Siong Chng School of Computer Engineering, Nanyang Technological

More information

Blind separation of instantaneous mixtures of dependent sources

Blind separation of instantaneous mixtures of dependent sources Blind separation of instantaneous mixtures of dependent sources Marc Castella and Pierre Comon GET/INT, UMR-CNRS 7, 9 rue Charles Fourier, 9 Évry Cedex, France marc.castella@int-evry.fr, CNRS, I3S, UMR

More information

1 Introduction Blind source separation (BSS) is a fundamental problem which is encountered in a variety of signal processing problems where multiple s

1 Introduction Blind source separation (BSS) is a fundamental problem which is encountered in a variety of signal processing problems where multiple s Blind Separation of Nonstationary Sources in Noisy Mixtures Seungjin CHOI x1 and Andrzej CICHOCKI y x Department of Electrical Engineering Chungbuk National University 48 Kaeshin-dong, Cheongju Chungbuk

More information

Analytical solution of the blind source separation problem using derivatives

Analytical solution of the blind source separation problem using derivatives Analytical solution of the blind source separation problem using derivatives Sebastien Lagrange 1,2, Luc Jaulin 2, Vincent Vigneron 1, and Christian Jutten 1 1 Laboratoire Images et Signaux, Institut National

More information

where A 2 IR m n is the mixing matrix, s(t) is the n-dimensional source vector (n» m), and v(t) is additive white noise that is statistically independ

where A 2 IR m n is the mixing matrix, s(t) is the n-dimensional source vector (n» m), and v(t) is additive white noise that is statistically independ BLIND SEPARATION OF NONSTATIONARY AND TEMPORALLY CORRELATED SOURCES FROM NOISY MIXTURES Seungjin CHOI x and Andrzej CICHOCKI y x Department of Electrical Engineering Chungbuk National University, KOREA

More information

An Iterative Blind Source Separation Method for Convolutive Mixtures of Images

An Iterative Blind Source Separation Method for Convolutive Mixtures of Images An Iterative Blind Source Separation Method for Convolutive Mixtures of Images Marc Castella and Jean-Christophe Pesquet Université de Marne-la-Vallée / UMR-CNRS 8049 5 bd Descartes, Champs-sur-Marne 77454

More information

Acoustic MIMO Signal Processing

Acoustic MIMO Signal Processing Yiteng Huang Jacob Benesty Jingdong Chen Acoustic MIMO Signal Processing With 71 Figures Ö Springer Contents 1 Introduction 1 1.1 Acoustic MIMO Signal Processing 1 1.2 Organization of the Book 4 Part I

More information

BLIND SEPARATION OF INSTANTANEOUS MIXTURES OF NON STATIONARY SOURCES

BLIND SEPARATION OF INSTANTANEOUS MIXTURES OF NON STATIONARY SOURCES BLIND SEPARATION OF INSTANTANEOUS MIXTURES OF NON STATIONARY SOURCES Dinh-Tuan Pham Laboratoire de Modélisation et Calcul URA 397, CNRS/UJF/INPG BP 53X, 38041 Grenoble cédex, France Dinh-Tuan.Pham@imag.fr

More information

Estimation of linear non-gaussian acyclic models for latent factors

Estimation of linear non-gaussian acyclic models for latent factors Estimation of linear non-gaussian acyclic models for latent factors Shohei Shimizu a Patrik O. Hoyer b Aapo Hyvärinen b,c a The Institute of Scientific and Industrial Research, Osaka University Mihogaoka

More information

BLIND DECONVOLUTION ALGORITHMS FOR MIMO-FIR SYSTEMS DRIVEN BY FOURTH-ORDER COLORED SIGNALS

BLIND DECONVOLUTION ALGORITHMS FOR MIMO-FIR SYSTEMS DRIVEN BY FOURTH-ORDER COLORED SIGNALS BLIND DECONVOLUTION ALGORITHMS FOR MIMO-FIR SYSTEMS DRIVEN BY FOURTH-ORDER COLORED SIGNALS M. Kawamoto 1,2, Y. Inouye 1, A. Mansour 2, and R.-W. Liu 3 1. Department of Electronic and Control Systems Engineering,

More information

Blind Identification of FIR Systems and Deconvolution of White Input Sequences

Blind Identification of FIR Systems and Deconvolution of White Input Sequences Blind Identification of FIR Systems and Deconvolution of White Input Sequences U. SOVERINI, P. CASTALDI, R. DIVERSI and R. GUIDORZI Dipartimento di Elettronica, Informatica e Sistemistica Università di

More information

Massoud BABAIE-ZADEH. Blind Source Separation (BSS) and Independent Componen Analysis (ICA) p.1/39

Massoud BABAIE-ZADEH. Blind Source Separation (BSS) and Independent Componen Analysis (ICA) p.1/39 Blind Source Separation (BSS) and Independent Componen Analysis (ICA) Massoud BABAIE-ZADEH Blind Source Separation (BSS) and Independent Componen Analysis (ICA) p.1/39 Outline Part I Part II Introduction

More information

Analytical Method for Blind Binary Signal Separation

Analytical Method for Blind Binary Signal Separation Analytical Method for Blind Binary Signal Separation Alle-Jan van der Veen Abstract The blind separation of multiple co-channel binary digital signals using an antenna array involves finding a factorization

More information

Acoustic Source Separation with Microphone Arrays CCNY

Acoustic Source Separation with Microphone Arrays CCNY Acoustic Source Separation with Microphone Arrays Lucas C. Parra Biomedical Engineering Department City College of New York CCNY Craig Fancourt Clay Spence Chris Alvino Montreal Workshop, Nov 6, 2004 Blind

More information

POLYNOMIAL SINGULAR VALUES FOR NUMBER OF WIDEBAND SOURCES ESTIMATION AND PRINCIPAL COMPONENT ANALYSIS

POLYNOMIAL SINGULAR VALUES FOR NUMBER OF WIDEBAND SOURCES ESTIMATION AND PRINCIPAL COMPONENT ANALYSIS POLYNOMIAL SINGULAR VALUES FOR NUMBER OF WIDEBAND SOURCES ESTIMATION AND PRINCIPAL COMPONENT ANALYSIS Russell H. Lambert RF and Advanced Mixed Signal Unit Broadcom Pasadena, CA USA russ@broadcom.com Marcel

More information

Blind Machine Separation Te-Won Lee

Blind Machine Separation Te-Won Lee Blind Machine Separation Te-Won Lee University of California, San Diego Institute for Neural Computation Blind Machine Separation Problem we want to solve: Single microphone blind source separation & deconvolution

More information

Robust extraction of specific signals with temporal structure

Robust extraction of specific signals with temporal structure Robust extraction of specific signals with temporal structure Zhi-Lin Zhang, Zhang Yi Computational Intelligence Laboratory, School of Computer Science and Engineering, University of Electronic Science

More information

REAL-TIME TIME-FREQUENCY BASED BLIND SOURCE SEPARATION. Scott Rickard, Radu Balan, Justinian Rosca. Siemens Corporate Research Princeton, NJ 08540

REAL-TIME TIME-FREQUENCY BASED BLIND SOURCE SEPARATION. Scott Rickard, Radu Balan, Justinian Rosca. Siemens Corporate Research Princeton, NJ 08540 REAL-TIME TIME-FREQUENCY BASED BLIND SOURCE SEPARATION Scott Rickard, Radu Balan, Justinian Rosca Siemens Corporate Research Princeton, NJ 84 fscott.rickard,radu.balan,justinian.roscag@scr.siemens.com

More information

TWO METHODS FOR ESTIMATING OVERCOMPLETE INDEPENDENT COMPONENT BASES. Mika Inki and Aapo Hyvärinen

TWO METHODS FOR ESTIMATING OVERCOMPLETE INDEPENDENT COMPONENT BASES. Mika Inki and Aapo Hyvärinen TWO METHODS FOR ESTIMATING OVERCOMPLETE INDEPENDENT COMPONENT BASES Mika Inki and Aapo Hyvärinen Neural Networks Research Centre Helsinki University of Technology P.O. Box 54, FIN-215 HUT, Finland ABSTRACT

More information

CCA BASED ALGORITHMS FOR BLIND EQUALIZATION OF FIR MIMO SYSTEMS

CCA BASED ALGORITHMS FOR BLIND EQUALIZATION OF FIR MIMO SYSTEMS CCA BASED ALGORITHMS FOR BLID EQUALIZATIO OF FIR MIMO SYSTEMS Javier Vía and Ignacio Santamaría Dept of Communications Engineering University of Cantabria 395 Santander, Cantabria, Spain E-mail: {jvia,nacho}@gtasdicomunicanes

More information

Tensor approach for blind FIR channel identification using 4th-order cumulants

Tensor approach for blind FIR channel identification using 4th-order cumulants Tensor approach for blind FIR channel identification using 4th-order cumulants Carlos Estêvão R Fernandes Gérard Favier and João Cesar M Mota contact: cfernand@i3s.unice.fr June 8, 2006 Outline 1. HOS

More information

ROBUSTNESS OF PARAMETRIC SOURCE DEMIXING IN ECHOIC ENVIRONMENTS. Radu Balan, Justinian Rosca, Scott Rickard

ROBUSTNESS OF PARAMETRIC SOURCE DEMIXING IN ECHOIC ENVIRONMENTS. Radu Balan, Justinian Rosca, Scott Rickard ROBUSTNESS OF PARAMETRIC SOURCE DEMIXING IN ECHOIC ENVIRONMENTS Radu Balan, Justinian Rosca, Scott Rickard Siemens Corporate Research Multimedia and Video Technology Princeton, NJ 5 fradu.balan,justinian.rosca,scott.rickardg@scr.siemens.com

More information

Blind Extraction of Singularly Mixed Source Signals

Blind Extraction of Singularly Mixed Source Signals IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL 11, NO 6, NOVEMBER 2000 1413 Blind Extraction of Singularly Mixed Source Signals Yuanqing Li, Jun Wang, Senior Member, IEEE, and Jacek M Zurada, Fellow, IEEE Abstract

More information

NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION. M. Schwab, P. Noll, and T. Sikora. Technical University Berlin, Germany Communication System Group

NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION. M. Schwab, P. Noll, and T. Sikora. Technical University Berlin, Germany Communication System Group NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION M. Schwab, P. Noll, and T. Sikora Technical University Berlin, Germany Communication System Group Einsteinufer 17, 1557 Berlin (Germany) {schwab noll

More information

PROPERTIES OF THE EMPIRICAL CHARACTERISTIC FUNCTION AND ITS APPLICATION TO TESTING FOR INDEPENDENCE. Noboru Murata

PROPERTIES OF THE EMPIRICAL CHARACTERISTIC FUNCTION AND ITS APPLICATION TO TESTING FOR INDEPENDENCE. Noboru Murata ' / PROPERTIES OF THE EMPIRICAL CHARACTERISTIC FUNCTION AND ITS APPLICATION TO TESTING FOR INDEPENDENCE Noboru Murata Waseda University Department of Electrical Electronics and Computer Engineering 3--

More information

GENERALIZED DEFLATION ALGORITHMS FOR THE BLIND SOURCE-FACTOR SEPARATION OF MIMO-FIR CHANNELS. Mitsuru Kawamoto 1,2 and Yujiro Inouye 1

GENERALIZED DEFLATION ALGORITHMS FOR THE BLIND SOURCE-FACTOR SEPARATION OF MIMO-FIR CHANNELS. Mitsuru Kawamoto 1,2 and Yujiro Inouye 1 GENERALIZED DEFLATION ALGORITHMS FOR THE BLIND SOURCE-FACTOR SEPARATION OF MIMO-FIR CHANNELS Mitsuru Kawamoto,2 and Yuiro Inouye. Dept. of Electronic and Control Systems Engineering, Shimane University,

More information

Simultaneous Diagonalization in the Frequency Domain (SDIF) for Source Separation

Simultaneous Diagonalization in the Frequency Domain (SDIF) for Source Separation Simultaneous Diagonalization in the Frequency Domain (SDIF) for Source Separation Hsiao-Chun Wu and Jose C. Principe Computational Neuro-Engineering Laboratory Department of Electrical and Computer Engineering

More information

Blind Channel Identification in (2 1) Alamouti Coded Systems Based on Maximizing the Eigenvalue Spread of Cumulant Matrices

Blind Channel Identification in (2 1) Alamouti Coded Systems Based on Maximizing the Eigenvalue Spread of Cumulant Matrices Blind Channel Identification in (2 1) Alamouti Coded Systems Based on Maximizing the Eigenvalue Spread of Cumulant Matrices Héctor J. Pérez-Iglesias 1, Daniel Iglesia 1, Adriana Dapena 1, and Vicente Zarzoso

More information

A UNIFIED PRESENTATION OF BLIND SEPARATION METHODS FOR CONVOLUTIVE MIXTURES USING BLOCK-DIAGONALIZATION

A UNIFIED PRESENTATION OF BLIND SEPARATION METHODS FOR CONVOLUTIVE MIXTURES USING BLOCK-DIAGONALIZATION A UNIFIED PRESENTATION OF BLIND SEPARATION METHODS FOR CONVOLUTIVE MIXTURES USING BLOCK-DIAGONALIZATION Cédric Févotte and Christian Doncarli Institut de Recherche en Communications et Cybernétique de

More information

An Improved Cumulant Based Method for Independent Component Analysis

An Improved Cumulant Based Method for Independent Component Analysis An Improved Cumulant Based Method for Independent Component Analysis Tobias Blaschke and Laurenz Wiskott Institute for Theoretical Biology Humboldt University Berlin Invalidenstraße 43 D - 0 5 Berlin Germany

More information

Semi-Blind approaches to source separation: introduction to the special session

Semi-Blind approaches to source separation: introduction to the special session Semi-Blind approaches to source separation: introduction to the special session Massoud BABAIE-ZADEH 1 Christian JUTTEN 2 1- Sharif University of Technology, Tehran, IRAN 2- Laboratory of Images and Signals

More information

On the Estimation of the Mixing Matrix for Underdetermined Blind Source Separation in an Arbitrary Number of Dimensions

On the Estimation of the Mixing Matrix for Underdetermined Blind Source Separation in an Arbitrary Number of Dimensions On the Estimation of the Mixing Matrix for Underdetermined Blind Source Separation in an Arbitrary Number of Dimensions Luis Vielva 1, Ignacio Santamaría 1,Jesús Ibáñez 1, Deniz Erdogmus 2,andJoséCarlosPríncipe

More information

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 53, NO. 12, DECEMBER

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 53, NO. 12, DECEMBER IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 53, NO 12, DECEMBER 2005 4429 Group Decorrelation Enhanced Subspace Method for Identifying FIR MIMO Channels Driven by Unknown Uncorrelated Colored Sources Senjian

More information

Blind Source Separation with a Time-Varying Mixing Matrix

Blind Source Separation with a Time-Varying Mixing Matrix Blind Source Separation with a Time-Varying Mixing Matrix Marcus R DeYoung and Brian L Evans Department of Electrical and Computer Engineering The University of Texas at Austin 1 University Station, Austin,

More information

Independent Component Analysis

Independent Component Analysis Independent Component Analysis James V. Stone November 4, 24 Sheffield University, Sheffield, UK Keywords: independent component analysis, independence, blind source separation, projection pursuit, complexity

More information

Fundamentals of Principal Component Analysis (PCA), Independent Component Analysis (ICA), and Independent Vector Analysis (IVA)

Fundamentals of Principal Component Analysis (PCA), Independent Component Analysis (ICA), and Independent Vector Analysis (IVA) Fundamentals of Principal Component Analysis (PCA),, and Independent Vector Analysis (IVA) Dr Mohsen Naqvi Lecturer in Signal and Information Processing, School of Electrical and Electronic Engineering,

More information

Improved PARAFAC based Blind MIMO System Estimation

Improved PARAFAC based Blind MIMO System Estimation Improved PARAFAC based Blind MIMO System Estimation Yuanning Yu, Athina P. Petropulu Department of Electrical and Computer Engineering Drexel University, Philadelphia, PA, 19104, USA This work has been

More information

Co-prime Arrays with Reduced Sensors (CARS) for Direction-of-Arrival Estimation

Co-prime Arrays with Reduced Sensors (CARS) for Direction-of-Arrival Estimation Co-prime Arrays with Reduced Sensors (CARS) for Direction-of-Arrival Estimation Mingyang Chen 1,LuGan and Wenwu Wang 1 1 Department of Electrical and Electronic Engineering, University of Surrey, U.K.

More information

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 53, NO. 8, AUGUST Nonlinear Adaptive Blind Whitening for MIMO Channels

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 53, NO. 8, AUGUST Nonlinear Adaptive Blind Whitening for MIMO Channels IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 53, NO 8, AUGUST 2005 2635 Nonlinear Adaptive Blind Whitening for MIMO Channels Han-Fu Chen, Fellow, IEEE, Xi-Ren Cao, Fellow, IEEE, Hai-Tao Fang, Member, IEEE,

More information

CONTROL SYSTEMS ANALYSIS VIA BLIND SOURCE DECONVOLUTION. Kenji Sugimoto and Yoshito Kikkawa

CONTROL SYSTEMS ANALYSIS VIA BLIND SOURCE DECONVOLUTION. Kenji Sugimoto and Yoshito Kikkawa CONTROL SYSTEMS ANALYSIS VIA LIND SOURCE DECONVOLUTION Kenji Sugimoto and Yoshito Kikkawa Nara Institute of Science and Technology Graduate School of Information Science 896-5 Takayama-cho, Ikoma-city,

More information

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 53, NO. 11, NOVEMBER Blind MIMO Identification Using the Second Characteristic Function

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 53, NO. 11, NOVEMBER Blind MIMO Identification Using the Second Characteristic Function IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 53, NO 11, NOVEMBER 2005 4067 Blind MIMO Identification Using the Second Characteristic Function Eran Eidinger and Arie Yeredor, Senior Member, IEEE Abstract

More information

BLIND SEPARATION OF INSTANTANEOUS MIXTURES OF NON STATIONARY SOURCES

BLIND SEPARATION OF INSTANTANEOUS MIXTURES OF NON STATIONARY SOURCES BLIND SEPARATION OF INSTANTANEOUS MIXTURES OF NON STATIONARY SOURCES Dinh-Tuan Pham Laboratoire de Modélisation et Calcul URA 397, CNRS/UJF/INPG BP 53X, 38041 Grenoble cédex, France Dinh-Tuan.Pham@imag.fr

More information

Improved system blind identification based on second-order cyclostationary statistics: A group delay approach

Improved system blind identification based on second-order cyclostationary statistics: A group delay approach SaÅdhanaÅ, Vol. 25, Part 2, April 2000, pp. 85±96. # Printed in India Improved system blind identification based on second-order cyclostationary statistics: A group delay approach P V S GIRIDHAR 1 and

More information

A Generalization of Blind Source Separation Algorithms for Convolutive Mixtures Based on Second-Order Statistics

A Generalization of Blind Source Separation Algorithms for Convolutive Mixtures Based on Second-Order Statistics 120 IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL 13, NO 1, JANUARY 2005 A Generalization of Blind Source Separation Algorithms for Convolutive Mixtures Based on Second-Order Statistics Herbert

More information

Sound Source Tracking Using Microphone Arrays

Sound Source Tracking Using Microphone Arrays Sound Source Tracking Using Microphone Arrays WANG PENG and WEE SER Center for Signal Processing School of Electrical & Electronic Engineering Nanayang Technological Univerisy SINGAPORE, 639798 Abstract:

More information

On the Behavior of Information Theoretic Criteria for Model Order Selection

On the Behavior of Information Theoretic Criteria for Model Order Selection IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 49, NO. 8, AUGUST 2001 1689 On the Behavior of Information Theoretic Criteria for Model Order Selection Athanasios P. Liavas, Member, IEEE, and Phillip A. Regalia,

More information

Vandermonde-form Preserving Matrices And The Generalized Signal Richness Preservation Problem

Vandermonde-form Preserving Matrices And The Generalized Signal Richness Preservation Problem Vandermonde-form Preserving Matrices And The Generalized Signal Richness Preservation Problem Borching Su Department of Electrical Engineering California Institute of Technology Pasadena, California 91125

More information

Research Letter An Algorithm to Generate Representations of System Identification Errors

Research Letter An Algorithm to Generate Representations of System Identification Errors Research Letters in Signal Processing Volume 008, Article ID 5991, 4 pages doi:10.1155/008/5991 Research Letter An Algorithm to Generate Representations of System Identification Errors Wancheng Zhang and

More information

SPARSE signal representations have gained popularity in recent

SPARSE signal representations have gained popularity in recent 6958 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 57, NO. 10, OCTOBER 2011 Blind Compressed Sensing Sivan Gleichman and Yonina C. Eldar, Senior Member, IEEE Abstract The fundamental principle underlying

More information

Using Kernel PCA for Initialisation of Variational Bayesian Nonlinear Blind Source Separation Method

Using Kernel PCA for Initialisation of Variational Bayesian Nonlinear Blind Source Separation Method Using Kernel PCA for Initialisation of Variational Bayesian Nonlinear Blind Source Separation Method Antti Honkela 1, Stefan Harmeling 2, Leo Lundqvist 1, and Harri Valpola 1 1 Helsinki University of Technology,

More information

Distinguishing Causes from Effects using Nonlinear Acyclic Causal Models

Distinguishing Causes from Effects using Nonlinear Acyclic Causal Models JMLR Workshop and Conference Proceedings 6:17 164 NIPS 28 workshop on causality Distinguishing Causes from Effects using Nonlinear Acyclic Causal Models Kun Zhang Dept of Computer Science and HIIT University

More information

Blind channel deconvolution of real world signals using source separation techniques

Blind channel deconvolution of real world signals using source separation techniques Blind channel deconvolution of real world signals using source separation techniques Jordi Solé-Casals 1, Enric Monte-Moreno 2 1 Signal Processing Group, University of Vic, Sagrada Família 7, 08500, Vic

More information

Independent Component Analysis. Contents

Independent Component Analysis. Contents Contents Preface xvii 1 Introduction 1 1.1 Linear representation of multivariate data 1 1.1.1 The general statistical setting 1 1.1.2 Dimension reduction methods 2 1.1.3 Independence as a guiding principle

More information

Efficient Semi-Blind Channel Estimation and Equalization Based on a Parametric Channel Representation

Efficient Semi-Blind Channel Estimation and Equalization Based on a Parametric Channel Representation Efficient Semi-Blind Channel Estimation and Equalization Based on a Parametric Channel Representation Presenter: Kostas Berberidis University of Patras Computer Engineering & Informatics Department Signal

More information

Title without the persistently exciting c. works must be obtained from the IEE

Title without the persistently exciting c.   works must be obtained from the IEE Title Exact convergence analysis of adapt without the persistently exciting c Author(s) Sakai, H; Yang, JM; Oka, T Citation IEEE TRANSACTIONS ON SIGNAL 55(5): 2077-2083 PROCESS Issue Date 2007-05 URL http://hdl.handle.net/2433/50544

More information

MULTICHANNEL SIGNAL PROCESSING USING SPATIAL RANK COVARIANCE MATRICES

MULTICHANNEL SIGNAL PROCESSING USING SPATIAL RANK COVARIANCE MATRICES MULTICHANNEL SIGNAL PROCESSING USING SPATIAL RANK COVARIANCE MATRICES S. Visuri 1 H. Oja V. Koivunen 1 1 Signal Processing Lab. Dept. of Statistics Tampere Univ. of Technology University of Jyväskylä P.O.

More information

Blind Source Separation Using Second-Order Cyclostationary Statistics

Blind Source Separation Using Second-Order Cyclostationary Statistics 694 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 49, NO. 4, APRIL 2001 Blind Source Separation Using Second-Order Cyclostationary Statistics Karim Abed-Meraim, Yong Xiang, Jonathan H. Manton, Yingbo Hua,

More information

BLIND system identification (BSI) is one of the fundamental

BLIND system identification (BSI) is one of the fundamental SUBMITTED TO IEEE SIGNAL PROCEING LETTERS, JANUARY 017 1 Structure-Based Subspace Method for Multi-Channel Blind System Identification Qadri Mayyala, Student Member, IEEE, Karim Abed-Meraim, Senior Member,

More information

IN HIGH-SPEED digital communications, the channel often

IN HIGH-SPEED digital communications, the channel often IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 45, NO 9, SEPTEMBER 1997 2277 Blind Fractionally Spaced Equalization of Noisy FIR Channels: Direct and Adaptive Solutions Georgios B Giannakis, Fellow, IEEE,

More information

Independent Component Analysis and Unsupervised Learning. Jen-Tzung Chien

Independent Component Analysis and Unsupervised Learning. Jen-Tzung Chien Independent Component Analysis and Unsupervised Learning Jen-Tzung Chien TABLE OF CONTENTS 1. Independent Component Analysis 2. Case Study I: Speech Recognition Independent voices Nonparametric likelihood

More information

Single Channel Signal Separation Using MAP-based Subspace Decomposition

Single Channel Signal Separation Using MAP-based Subspace Decomposition Single Channel Signal Separation Using MAP-based Subspace Decomposition Gil-Jin Jang, Te-Won Lee, and Yung-Hwan Oh 1 Spoken Language Laboratory, Department of Computer Science, KAIST 373-1 Gusong-dong,

More information

A METHOD OF ICA IN TIME-FREQUENCY DOMAIN

A METHOD OF ICA IN TIME-FREQUENCY DOMAIN A METHOD OF ICA IN TIME-FREQUENCY DOMAIN Shiro Ikeda PRESTO, JST Hirosawa 2-, Wako, 35-98, Japan Shiro.Ikeda@brain.riken.go.jp Noboru Murata RIKEN BSI Hirosawa 2-, Wako, 35-98, Japan Noboru.Murata@brain.riken.go.jp

More information

Wavelet de-noising for blind source separation in noisy mixtures.

Wavelet de-noising for blind source separation in noisy mixtures. Wavelet for blind source separation in noisy mixtures. Bertrand Rivet 1, Vincent Vigneron 1, Anisoara Paraschiv-Ionescu 2 and Christian Jutten 1 1 Institut National Polytechnique de Grenoble. Laboratoire

More information

Blind signal processing algorithms

Blind signal processing algorithms 12th Int. Workshop on Systems, Signals & Image Processing, 22-24 September 2005, Chalkida, Greece 105 Blind signal processing algorithms Athanasios Margaris and Efthimios Kotsialos Department of Applied

More information

TRINICON: A Versatile Framework for Multichannel Blind Signal Processing

TRINICON: A Versatile Framework for Multichannel Blind Signal Processing TRINICON: A Versatile Framework for Multichannel Blind Signal Processing Herbert Buchner, Robert Aichner, Walter Kellermann {buchner,aichner,wk}@lnt.de Telecommunications Laboratory University of Erlangen-Nuremberg

More information

LINEAR parametric models have found widespread use

LINEAR parametric models have found widespread use 3084 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 45, NO 12, DECEMBER 1997 Subspace Methods for Blind Estimation of Time-Varying FIR Channels Michail K Tsatsanis and Georgios B Giannakis, Fellow, IEEE Abstract

More information

Verification of contribution separation technique for vehicle interior noise using only response signals

Verification of contribution separation technique for vehicle interior noise using only response signals Verification of contribution separation technique for vehicle interior noise using only response signals Tomohiro HIRANO 1 ; Junji YOSHIDA 1 1 Osaka Institute of Technology, Japan ABSTRACT In this study,

More information

1 Introduction Independent component analysis (ICA) [10] is a statistical technique whose main applications are blind source separation, blind deconvo

1 Introduction Independent component analysis (ICA) [10] is a statistical technique whose main applications are blind source separation, blind deconvo The Fixed-Point Algorithm and Maximum Likelihood Estimation for Independent Component Analysis Aapo Hyvarinen Helsinki University of Technology Laboratory of Computer and Information Science P.O.Box 5400,

More information

On the INFOMAX algorithm for blind signal separation

On the INFOMAX algorithm for blind signal separation University of Wollongong Research Online Faculty of Informatics - Papers (Archive) Faculty of Engineering and Information Sciences 2000 On the INFOMAX algorithm for blind signal separation Jiangtao Xi

More information

Blind Instantaneous Noisy Mixture Separation with Best Interference-plus-noise Rejection

Blind Instantaneous Noisy Mixture Separation with Best Interference-plus-noise Rejection Blind Instantaneous Noisy Mixture Separation with Best Interference-plus-noise Rejection Zbyněk Koldovský 1,2 and Petr Tichavský 1 1 Institute of Information Theory and Automation, Pod vodárenskou věží

More information

Impulsive Noise Filtering In Biomedical Signals With Application of New Myriad Filter

Impulsive Noise Filtering In Biomedical Signals With Application of New Myriad Filter BIOSIGAL 21 Impulsive oise Filtering In Biomedical Signals With Application of ew Myriad Filter Tomasz Pander 1 1 Division of Biomedical Electronics, Institute of Electronics, Silesian University of Technology,

More information

Independent Component Analysis and Unsupervised Learning

Independent Component Analysis and Unsupervised Learning Independent Component Analysis and Unsupervised Learning Jen-Tzung Chien National Cheng Kung University TABLE OF CONTENTS 1. Independent Component Analysis 2. Case Study I: Speech Recognition Independent

More information

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 54, NO. 2, FEBRUARY

IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 54, NO. 2, FEBRUARY IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 54, NO 2, FEBRUARY 2006 423 Underdetermined Blind Source Separation Based on Sparse Representation Yuanqing Li, Shun-Ichi Amari, Fellow, IEEE, Andrzej Cichocki,

More information

Statistical and Adaptive Signal Processing

Statistical and Adaptive Signal Processing r Statistical and Adaptive Signal Processing Spectral Estimation, Signal Modeling, Adaptive Filtering and Array Processing Dimitris G. Manolakis Massachusetts Institute of Technology Lincoln Laboratory

More information

Chapter 2 Fundamentals of Adaptive Filter Theory

Chapter 2 Fundamentals of Adaptive Filter Theory Chapter 2 Fundamentals of Adaptive Filter Theory In this chapter we will treat some fundamentals of the adaptive filtering theory highlighting the system identification problem We will introduce a signal

More information

Time-domain representations

Time-domain representations Time-domain representations Speech Processing Tom Bäckström Aalto University Fall 2016 Basics of Signal Processing in the Time-domain Time-domain signals Before we can describe speech signals or modelling

More information

Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi

Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi Communication Engineering Prof. Surendra Prasad Department of Electrical Engineering Indian Institute of Technology, Delhi Lecture - 41 Pulse Code Modulation (PCM) So, if you remember we have been talking

More information

Distinguishing Causes from Effects using Nonlinear Acyclic Causal Models

Distinguishing Causes from Effects using Nonlinear Acyclic Causal Models Distinguishing Causes from Effects using Nonlinear Acyclic Causal Models Kun Zhang Dept of Computer Science and HIIT University of Helsinki 14 Helsinki, Finland kun.zhang@cs.helsinki.fi Aapo Hyvärinen

More information

Sparse filter models for solving permutation indeterminacy in convolutive blind source separation

Sparse filter models for solving permutation indeterminacy in convolutive blind source separation Sparse filter models for solving permutation indeterminacy in convolutive blind source separation Prasad Sudhakar, Rémi Gribonval To cite this version: Prasad Sudhakar, Rémi Gribonval. Sparse filter models

More information

Unsupervised Learning with Permuted Data

Unsupervised Learning with Permuted Data Unsupervised Learning with Permuted Data Sergey Kirshner skirshne@ics.uci.edu Sridevi Parise sparise@ics.uci.edu Padhraic Smyth smyth@ics.uci.edu School of Information and Computer Science, University

More information

STATISTICAL MODELLING OF MULTICHANNEL BLIND SYSTEM IDENTIFICATION ERRORS. Felicia Lim, Patrick A. Naylor

STATISTICAL MODELLING OF MULTICHANNEL BLIND SYSTEM IDENTIFICATION ERRORS. Felicia Lim, Patrick A. Naylor STTISTICL MODELLING OF MULTICHNNEL BLIND SYSTEM IDENTIFICTION ERRORS Felicia Lim, Patrick. Naylor Dept. of Electrical and Electronic Engineering, Imperial College London, UK {felicia.lim6, p.naylor}@imperial.ac.uk

More information

Computing Consecutive-Type Reliabilities Non-Recursively

Computing Consecutive-Type Reliabilities Non-Recursively IEEE TRANSACTIONS ON RELIABILITY, VOL. 52, NO. 3, SEPTEMBER 2003 367 Computing Consecutive-Type Reliabilities Non-Recursively Galit Shmueli Abstract The reliability of consecutive-type systems has been

More information

BLIND SOURCE SEPARATION TECHNIQUES ANOTHER WAY OF DOING OPERATIONAL MODAL ANALYSIS

BLIND SOURCE SEPARATION TECHNIQUES ANOTHER WAY OF DOING OPERATIONAL MODAL ANALYSIS BLIND SOURCE SEPARATION TECHNIQUES ANOTHER WAY OF DOING OPERATIONAL MODAL ANALYSIS F. Poncelet, Aerospace and Mech. Eng. Dept., University of Liege, Belgium G. Kerschen, Aerospace and Mech. Eng. Dept.,

More information

Underdetermined Instantaneous Audio Source Separation via Local Gaussian Modeling

Underdetermined Instantaneous Audio Source Separation via Local Gaussian Modeling Underdetermined Instantaneous Audio Source Separation via Local Gaussian Modeling Emmanuel Vincent, Simon Arberet, and Rémi Gribonval METISS Group, IRISA-INRIA Campus de Beaulieu, 35042 Rennes Cedex, France

More information

798 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 10, OCTOBER 1997

798 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL. 44, NO. 10, OCTOBER 1997 798 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II: ANALOG AND DIGITAL SIGNAL PROCESSING, VOL 44, NO 10, OCTOBER 1997 Stochastic Analysis of the Modulator Differential Pulse Code Modulator Rajesh Sharma,

More information

STUDY ON METHODS FOR COMPUTER-AIDED TOOTH SHADE DETERMINATION

STUDY ON METHODS FOR COMPUTER-AIDED TOOTH SHADE DETERMINATION INTERNATIONAL JOURNAL OF INFORMATION AND SYSTEMS SCIENCES Volume 5, Number 3-4, Pages 351 358 c 2009 Institute for Scientific Computing and Information STUDY ON METHODS FOR COMPUTER-AIDED TOOTH SHADE DETERMINATION

More information

Beamforming Using the Fractional Fourier Transform

Beamforming Using the Fractional Fourier Transform IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 51, NO. 6, JUNE 2003 1663 Then, using similar arguments to the ones in [3], it can be shown that [3, Th. 1] holds. Moreover, [3, Th. 2] holds with the following

More information

Single Channel Music Sound Separation Based on Spectrogram Decomposition and Note Classification

Single Channel Music Sound Separation Based on Spectrogram Decomposition and Note Classification Single Channel Music Sound Separation Based on Spectrogram Decomposition and Note Classification Hafiz Mustafa and Wenwu Wang Centre for Vision, Speech and Signal Processing (CVSSP) University of Surrey,

More information

x 1 (t) Spectrogram t s

x 1 (t) Spectrogram t s A METHOD OF ICA IN TIME-FREQUENCY DOMAIN Shiro Ikeda PRESTO, JST Hirosawa 2-, Wako, 35-98, Japan Shiro.Ikeda@brain.riken.go.jp Noboru Murata RIKEN BSI Hirosawa 2-, Wako, 35-98, Japan Noboru.Murata@brain.riken.go.jp

More information

FAST AND EFFECTIVE MODEL ORDER SELECTION METHOD TO DETERMINE THE NUMBER OF SOURCES IN A LINEAR TRANSFORMATION MODEL

FAST AND EFFECTIVE MODEL ORDER SELECTION METHOD TO DETERMINE THE NUMBER OF SOURCES IN A LINEAR TRANSFORMATION MODEL FAST AND EFFECTIVE MODEL ORDER SELECTION METHOD TO DETERMINE THE NUMBER OF SOURCES IN A LINEAR TRANSFORMATION MODEL Fengyu Cong 1, Asoke K Nandi 1,2, Zhaoshui He 3, Andrzej Cichocki 4, Tapani Ristaniemi

More information

Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation

Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation Blind Spectral-GMM Estimation for Underdetermined Instantaneous Audio Source Separation Simon Arberet 1, Alexey Ozerov 2, Rémi Gribonval 1, and Frédéric Bimbot 1 1 METISS Group, IRISA-INRIA Campus de Beaulieu,

More information

Covariance smoothing and consistent Wiener filtering for artifact reduction in audio source separation

Covariance smoothing and consistent Wiener filtering for artifact reduction in audio source separation Covariance smoothing and consistent Wiener filtering for artifact reduction in audio source separation Emmanuel Vincent METISS Team Inria Rennes - Bretagne Atlantique E. Vincent (Inria) Artifact reduction

More information

Cooperative Communication with Feedback via Stochastic Approximation

Cooperative Communication with Feedback via Stochastic Approximation Cooperative Communication with Feedback via Stochastic Approximation Utsaw Kumar J Nicholas Laneman and Vijay Gupta Department of Electrical Engineering University of Notre Dame Email: {ukumar jnl vgupta}@ndedu

More information

MIMO instantaneous blind identification based on second-order temporal structure

MIMO instantaneous blind identification based on second-order temporal structure MIMO instantaneous blind identification based on second-order temporal structure Citation for published version (APA): Laar, van de, J, Moonen, M, & Sommen, P C W (2008) MIMO instantaneous blind identification

More information

Comparative Performance Analysis of Three Algorithms for Principal Component Analysis

Comparative Performance Analysis of Three Algorithms for Principal Component Analysis 84 R. LANDQVIST, A. MOHAMMED, COMPARATIVE PERFORMANCE ANALYSIS OF THR ALGORITHMS Comparative Performance Analysis of Three Algorithms for Principal Component Analysis Ronnie LANDQVIST, Abbas MOHAMMED Dept.

More information

Blind Source Separation via Generalized Eigenvalue Decomposition

Blind Source Separation via Generalized Eigenvalue Decomposition Journal of Machine Learning Research 4 (2003) 1261-1269 Submitted 10/02; Published 12/03 Blind Source Separation via Generalized Eigenvalue Decomposition Lucas Parra Department of Biomedical Engineering

More information

On Information Maximization and Blind Signal Deconvolution

On Information Maximization and Blind Signal Deconvolution On Information Maximization and Blind Signal Deconvolution A Röbel Technical University of Berlin, Institute of Communication Sciences email: roebel@kgwtu-berlinde Abstract: In the following paper we investigate

More information