scispace - formally typeset
Search or ask a question

Showing papers in "IEEE Signal Processing Letters in 1995"


Journal ArticleDOI
TL;DR: This paper draws attention to a fundamental problem of multichannel AEC that concerns the nonunique nature of the estimated receiving room impulse responses.
Abstract: As teleconferencing systems evolve to an ever more lifelike and transparent audio/video medium, it will be necessary to incorporate multichannel audio, which at a minimum involves two channels, i.e., stereophonic sound. However, before full-duplex stereophonic teleconferencing can be deployed, the acoustic echo cancellation (AEC) problem must be solved in this regime. This paper draws attention to a fundamental problem of multichannel AEC that concerns the nonunique nature of the estimated receiving room impulse responses. Various potential solutions are discussed but are shown to be not entirely satisfactory. >

327 citations


Journal ArticleDOI
TL;DR: Generalized Gaussian and Laplacian source models are compared in discrete cosine transform (DCT) image coding and with block classification based on AC energy, the densities of the DCT coefficients are much closer to the LaPLacian or even the Gaussian.
Abstract: Generalized Gaussian and Laplacian source models are compared in discrete cosine transform (DCT) image coding. A difference in peak signal to noise ratio (PSNR) of at most 0.5 dB is observed for encoding different images. We also compare maximum likelihood estimation of the generalized Gaussian density parameters with a simpler method proposed by Mallat (1989). With block classification based on AC energy, the densities of the DCT coefficients are much closer to the Laplacian or even the Gaussian. >

145 citations


Journal ArticleDOI
TL;DR: Instantaneous signal operators /spl Upsi//sub k/(x)=x/spl dot/x/Sup (k-1)/-xx/sup (k)/ of integer orders k are proposed to measure the cross energy between a signal x and its derivatives.
Abstract: Instantaneous signal operators /spl Upsi//sub k/(x)=x/spl dot/x/sup (k-1)/-xx/sup (k)/ of integer orders k are proposed to measure the cross energy between a signal x and its derivatives. These higher order differential energy operators contain as a special case, for k=2, the Teager-Kaiser (1990) operator. When applied to (possibly modulated) sinusoids, they yield several new energy measurements useful for parameter estimation or AM-FM demodulation. Applying them to sampled signals involves replacing derivatives with differences that lead to several useful discrete energy operators defined on an extremely short window of samples. >

116 citations


Journal ArticleDOI
TL;DR: Using oversampling and the finite-alphabet property of digital communication signals, it is possible to blindly identify an FIR channel carrying a superposition of such signals, provided they have the same (known) period.
Abstract: Using oversampling and the finite-alphabet property of digital communication signals, it is possible to blindly identify an FIR channel carrying a superposition of such signals, provided they have the same (known) period, and certain rank conditions on the data and channel matrices are satisfied. In particular, this technique allows separation of finite alphabet signals, removal of intersymbol interference, and synchronization of the signals. An algorithm is proposed and tested on simulated data. >

111 citations


Journal ArticleDOI
TL;DR: This paper presents text-independent speaker identification results for varying speaker population sizes up to 630 speakers for both clean, wideband speech, and telephone speech using a system based on Gaussian mixture speaker models.
Abstract: This paper presents text-independent speaker identification results for varying speaker population sizes up to 630 speakers for both clean, wideband speech, and telephone speech. A system based on Gaussian mixture speaker models is used for speaker identification, and experiments are conducted on the TIMIT and NTIMIT databases. The TIMIT results show large population performance under near-ideal conditions, and the NTIMIT results show the corresponding accuracy loss due to telephone transmission. These are believed to be the first speaker identification experiments on the complete 630 speaker TIMIT and NTIMIT databases and the largest text-independent speaker identification task reported to date. Identification accuracies of 99.5 and 60.7% were achieved on the TIMIT and NTIMIT databases, respectively. >

111 citations


Journal ArticleDOI
Bin Yang1
TL;DR: An extension of the PASTd algorithm to both rank and signal subspace tracking that has a low computational complexity O(nr), where n is the input vector length, and r denotes the signal sub space dimension.
Abstract: In this letter, we present an extension of the PASTd algorithm to both rank and signal subspace tracking. It has a low computational complexity O(nr), where n is the input vector length, and r denotes the signal subspace dimension. Its performance in tracking time-varying direction of arrival is comparable with that of the expensive eigenvalue decomposition and more robust than the O(n/sup 2/) rank revealing URV updating algorithm proposed by Stewart. >

97 citations


Journal ArticleDOI
TL;DR: This letter proposes a method for blind separation of d co-channel BPSK signals arriving at an antenna array using computer simulations to study the performance of this method.
Abstract: In this letter, we propose a method for blind separation of d co-channel BPSK signals arriving at an antenna array. Our method involves two steps. In the first step, the received data vectors at the output of the array is grouped into 2/sup d/ clusters. In the second step, we assign the 2/sup d/ d-tuples with /spl plusmn/1 elements to these clusters in a consistent fashion. From the knowledge of the cluster to which a data vector belongs, we estimate the bits transmitted at that instant. Computer simulations are used to study the performance of our method. >

87 citations


Journal ArticleDOI
TL;DR: The authors address the problem of maximum likelihood estimation of dependence tree models with missing observations, using the expectation-maximization algorithm, which involves computing observation probabilities with an iterative "upward-downward" algorithm.
Abstract: A dependence tree is a model for the joint probability distribution of an n-dimensional random vector, which requires a relatively small number of free parameters by making Markov-like assumptions on the tree. The authors address the problem of maximum likelihood estimation of dependence tree models with missing observations, using the expectation-maximization algorithm. The solution involves computing observation probabilities with an iterative "upward-downward" algorithm, which is similar to an algorithm proposed for belief propagation in causal trees, a special case of Bayesian networks. >

75 citations


Journal ArticleDOI
TL;DR: Increasing the bandwidth of analysis filters relative to the synthesis filters is proposed to reduce the slow asymptotic convergence associated with oversampled systems.
Abstract: The motivation for adaptive filtering in subbands stems from two well-known problems in least-mean square full-band adaptive filtering. First, the convergence and tracking can be very slow if the input correlation matrix is ill conditioned, as in the case with speech input. Second, very high order adaptive filters are computationally expensive. One problem with adaptive filtering in subbands is the slow, asymptotic convergence associated with oversampled systems. Increasing the bandwidth of analysis filters relative to the synthesis filters is proposed to reduce the slow asymptotic convergence. The authors present experimental results illustrating the benefits of this modification. >

57 citations


Journal ArticleDOI
R.D. Poltmann1
TL;DR: It is shown in which way the delayed LMS (DLMS) algorithm can be transformed into the standard LMS algorithm at only slightly increased computational expense.
Abstract: For some applications of adaptive finite impulse response (FIR) filtering, the adaptation algorithm can be implemented only with a delay in the coefficient update. It is well known that this has an adverse effect on the convergence behavior of the algorithm. It is shown in which way the delayed LMS (DLMS) algorithm can be transformed into the standard LMS algorithm at only slightly increased computational expense.

55 citations


Journal ArticleDOI
T. Chen1, H.P. Graf1, Kuansan Wang2
TL;DR: The marriage of speech analysis and image processing can solve problems related to lip synchronization and speech information is utilized to improve the quality of audio-visual communications such as videotelephony and videoconferencing.
Abstract: We utilize speech information to improve the quality of audio-visual communications such as videotelephony and videoconferencing. In particular, the marriage of speech analysis and image processing can solve problems related to lip synchronization. We present a technique called speech-assisted frame-rate conversion. Demonstration sequences are presented. Other applications, including speech-assisted video coding, are outlined. >

Journal ArticleDOI
TL;DR: Clenshaw's recurrence formula provides a unified development for the recursive DCT and IDCT algorithms and applies to arbitrary length algorithms and are appropriate for VLSI implementation.
Abstract: Clenshaw's recurrence formula is used to derive recursive algorithms for the discrete cosine transform (DCT) and the inverse discrete cosine transform (IDCT). The recursive DCT algorithm presented requires one fewer delay element per coefficient and one fewer multiply operation per coefficient compared with two other proposed methods. Clenshaw's recurrence formula provides a unified development for the recursive DCT and IDCT algorithms. The recursive algorithms apply to arbitrary length algorithms and are appropriate for VLSI implementation. >

Journal ArticleDOI
F. Bimbot, R. Pieraccini1, E. Levin1, B. Atal1
TL;DR: A model that represents sentences as a concatenation of variable-length sequences of units and an algorithm for unsupervised estimation of the model parameters is presented and an approach is illustrated for the segmentation of sequences of letters into subword-like units.
Abstract: The conventional n-gram language model exploits dependencies between words and their fixed-length past. This letter presents a model that represents sentences as a concatenation of variable-length sequences of units and describes an algorithm for unsupervised estimation of the model parameters. The approach is illustrated for the segmentation of sequences of letters into subword-like units. It is evaluated as a language model on a corpus of transcribed spoken sentences. Multigrams can provide a significantly lower test set perplexity than n-gram models. >

Journal ArticleDOI
TL;DR: A nonuniform sampling series for nonbandlimited functions (duration-limited functions of bounded variation) together with error bounds and the bound on the approximation error, remarkably, does not depend on the distribution of the sampling points.
Abstract: We give a nonuniform sampling series for nonbandlimited functions (duration-limited functions of bounded variation), together with error bounds. The bound on the approximation error, remarkably, does not depend on the distribution of the sampling points. >

Journal ArticleDOI
TL;DR: An improved wavelet compression algorithm for ECG signals has been developed with the use of vector quantization on wavelet coefficients and preliminary results indicate that the proposed method excels over standard techniques for high fidelity compression.
Abstract: An improved wavelet compression algorithm for ECG signals has been developed with the use of vector quantization on wavelet coefficients. Vector quantization on scales of long duration and low dynamic range retains feature integrity of the ECG with a very low bit-per-sample rate. Preliminary results indicate that the proposed method excels over standard techniques for high fidelity compression. >

Journal ArticleDOI
V.J. Mathews1
TL;DR: This article presents a simple method for orthogonalizing correlated Gaussian input signals for identification of truncated Volterra systems of arbitrary order of nonlinearity P and memory length N.
Abstract: This article presents a simple method for orthogonalizing correlated Gaussian input signals for identification of truncated Volterra systems of arbitrary order of nonlinearity P and memory length N. The procedure requires a Gram-Schmidt orthogonalizer for a vector containing N elements and some nonlinear processing of the output elements of the Gram-Schmidt processor. However, the nonlinear processors do not depend on the statistics of the input signals and, consequently, are easy to design and implement. >

Journal ArticleDOI
TL;DR: The paper establishes the relationship between the two methods of higher order time-frequency analysis: the polynomial Wigner-Ville distribution (WVD) and the higher order WVD by using the projection-slice theorem.
Abstract: The paper establishes the relationship between the two methods of higher order time-frequency analysis: the polynomial Wigner-Ville distribution (WVD) and the higher order WVD. Using the projection-slice theorem, it is shown that the polynomial WVD represents a unique projection of the higher-order WVD from the time-multifrequency space to the time-frequency subspace. The implication of this relationship is investigated from the aspect of the analysis of multicomponent signals.

Journal ArticleDOI
TL;DR: The popular spatial smoothing technique is considered and it is shown via the covariance matrix eigenvalue analysis that the simple suboptimal rule for choosing of the subarray size exists in a practically important situation of two coherent equipower closely spaced sources.
Abstract: We consider the popular spatial smoothing technique and show via the covariance matrix eigenvalue analysis that the simple suboptimal rule for choosing of the subarray size exists in a practically important situation of two coherent equipower closely spaced sources. This rule has been derived by maximizing the distance between the signal subspace and the noise subspace eigenvalues of spatially smoothed covariance matrix and it does not require any a priori information about the signal source parameters. >

Journal ArticleDOI
TL;DR: This letter describes a CAD system for automatic implementation of FIR filters on Xilinx field programmable gate arrays (FPGA) given the frequency specifications, and the FPGA specific mapping techniques used to increase speed are discussed.
Abstract: This letter describes a CAD system for automatic implementation of FIR filters on Xilinx field programmable gate arrays (FPGA). Given the frequency specifications, this software package designs an FIR filter, optimizes the filter coefficients in the power of two coefficient space, and implements it on FPGA chips. The FPGA specific mapping techniques used to increase speed are discussed. The performance of the typical filters that were implemented is presented. >

Journal ArticleDOI
TL;DR: The methods of Cohen (1991) and of Baraniuk and Jones (see Proc. Acoust., Speech Signal Processing, ICASSP '93) are compared and their equivalence for variables that have the same commutator as time and frequency is shown.
Abstract: There has been considerable interest in the problem of joint representations for variables other than time and frequency. We compare the methods of Cohen (1991) and of Baraniuk and Jones (see Proc. IEEE Int. Conf. Acoust., Speech Signal Processing, ICASSP '93, p.320-323, vol.III) and show their equivalence for variables that have the same commutator as time and frequency. In addition, we report the following very general result. All pairs of variables connected by a unitary transformation have joint distributions that are functionally equivalent. >

Journal ArticleDOI
TL;DR: A Bayesian spectrum estimator of harmonic signals in Gaussian noise is derived based on the expected value of the theoretical signal spectrum over the joint posterior density function of the signal and noise parameters.
Abstract: A Bayesian spectrum estimator of harmonic signals in Gaussian noise is derived. It is based on the expected value of the theoretical signal spectrum over the joint posterior density function of the signal and noise parameters. Simulation results are provided that show its performance and comparison with MUSIC. >

Journal ArticleDOI
TL;DR: A sensor array processing algorithm for joint estimation of the directions of arrival and the corresponding angular velocities is presented, together with the Cramer-Rao bound (CRB) of the problem.
Abstract: A sensor array processing algorithm for joint estimation of the directions of arrival (DOAs) and the corresponding angular velocities is presented, together with the Cramer-Rao bound (CRB) of the problem. An embedded movement model circumvents the problems of biased estimates or high measurement rates that occur when conventional DOA estimation algorithms are applied to moving targets. >

Journal ArticleDOI
TL;DR: An automatic histogram specification method is proposed for image enhancement which meets the image enhancement requirements and the performance and effectiveness are shown by the analysis and the resultant image in comparison with other methods.
Abstract: An automatic histogram specification method is proposed for image enhancement. Parameterized fuzzy number (PFN) is adopted for the representation of image histogram. The desired histogram PFN is automatically constructed by the fuzzy addition of the tuning PFN and the original one. The tuning PFN is implied from the fuzzy inference rules. The mapping function is decided by the relations between the desired PFN and the original one. The effectual results are demonstrated by the mapping functions which meet the image enhancement requirements. The performance and effectiveness are shown by the analysis and the resultant image in comparison with other methods. >

Journal ArticleDOI
TL;DR: A class of multistage algorithms for carrying out incremental refinement of DFT and STFT approximations that rely almost exclusively on vector summation operations and can be designed to exhibit a variety of tradeoffs between improvement in approximation quality and computational cost per stage.
Abstract: We present a class of multistage algorithms for carrying out incremental refinement of DFT and STFT approximations. Each stage is designed to improve the previous stage's approximation in terms of frequency coverage, frequency resolution, and SNR. These algorithms rely almost exclusively on vector summation operations, and they can be designed to exhibit a variety of tradeoffs between improvement in approximation quality and computational cost per stage. The performance of incremental STFT refinement on real data serves to illustrate the relevance of such algorithms to application systems with dynamic real-time constraints. >

Journal ArticleDOI
TL;DR: A class of highly regular fast cyclic convolution algorithms, based on block pseudocirculant matrices, is obtained.
Abstract: Pseudocirculant matrices have been studied in the past in the context of FIR filtering, block filtering, polyphase networks and others. For completeness, their relation to cyclic convolution, stride permutations, circulant matrices, and to certain permutations of the Fourier matrix is explicitly established in this work. Within this process, a class of highly regular fast cyclic convolution algorithms, based on block pseudocirculant matrices, is obtained. >

Journal ArticleDOI
TL;DR: This work has calculated the distortion associated with Z, E/sub 8/, and Leech (1967) lattice quantization for coding of generalized Gaussian sources and shows that for low bit rates, the Z lattice offers both the best performance and the lowest implementational complexity.
Abstract: In lattice vector quantization, the distortion associated with a given lattice is often expressed in terms of the G number, which is a measure of the mean square error per dimension generated by quantization of a uniform source. Subband image coefficients, however, are best modeled by a generalized Gaussian distribution, leading to distortion characteristics that are quite different from those encountered for uniform, Laplacian, or Gaussian sources. We have calculated the distortion associated with Z, E/sub 8/, and Leech (1967) lattice quantization for coding of generalized Gaussian sources and show that for low bit rates, the Z lattice offers both the best performance and the lowest implementational complexity. >

Journal ArticleDOI
TL;DR: A simple proof of an autoregressive constraint on the input is sufficient to ensure stability of the furnished model for multivariable system estimation is provided.
Abstract: Least-squares equation-error models are widely used as a simple means of estimating an input-output transfer function in a system identification context.. Although the models furnished by the least-squares method are not always stable, some recent works have shown that an autoregressive constraint on the input is sufficient to ensure stability of the furnished model. Here we provide a simple proof of this property for multivariable system estimation. >

Journal ArticleDOI
TL;DR: This letter shows how to incorporate boundary values into a multidimensional wave digital filter, using a lossy transmission line of finite extent with given terminations as an example.
Abstract: The application of multidimensional wave digital filter principles to the numerical solution of partial differential equations has been introduced in various recent publications. However, the examples presented so far do not explicitly consider boundary conditions. This letter shows how to incorporate boundary values into a multidimensional wave digital filter. A lossy transmission line of finite extent with given terminations is chosen as an example. >

Journal ArticleDOI
TL;DR: A new method to stabilize the algorithm is introduced that does not depend on the characteristics of the excitation signal, and by inserting leakage factors in the filter equations, a steady contraction of error components is reached.
Abstract: A derivation of fundamental fast transversal filter equations comes to a 7M fast RLS algorithm for system identification. Yet, this algorithm is not stable due to numerical errors resulting from implementation on a hardware with finite precision arithmetic. The known stabilization methods with feedback of numerical errors require a judicious choice of the feedback gains and depend on the properties of the exciting sequence, as simulations have shown. A new method to stabilize the algorithm is introduced that does not depend on the characteristics of the excitation signal. By inserting leakage factors in the filter equations, a steady contraction of error components is reached. >

Journal ArticleDOI
TL;DR: This letter presents a novel application of Peano scanning to expedite the partial distance search (PDS) when coding image blocks using vector quantization (VQ) and computes and stores a set of sorted tables of PS values of selected feature vectors associated with the codevectors.
Abstract: This letter presents a novel application of Peano scanning (PS) to expedite the partial distance search (PDS) when coding image blocks using vector quantization (VQ). It computes and stores a set of sorted tables of PS values of selected feature vectors associated with the codevectors. A PS acts as a transform from a higher dimension to a single dimension while preserving neighborhood adjacency. Coding of an input block is accomplished by finding the closest PS value to that of the input and its corresponding table. PDS proceeds from the codevector of this PS value. Computational complexity is greatly reduced with the PS value easily obtained in hardware and real time. >