Efficiently solving a system identification problem represents an important step in numerous impo... more Efficiently solving a system identification problem represents an important step in numerous important applications. In this framework, some of the most popular solutions rely on the Wiener filter, which is widely used in practice. Moreover, it also represents a benchmark for other related optimization problems. In this paper, new insights into the regularization of the Wiener filter are provided, which is a must in real-world scenarios. A proper regularization technique is of great importance, especially in challenging conditions, e.g., when operating in noisy environments and/or when only a low quantity of data is available for the estimation of the statistics. Different regularization methods are investigated in this paper, including several new solutions that fit very well for the identification of sparse and low-rank systems. Experimental results support the theoretical developments and indicate the efficiency of the proposed techniques.
Due to its fast convergence rate, the recursive least-squares (RLS) algorithm is very popular in ... more Due to its fast convergence rate, the recursive least-squares (RLS) algorithm is very popular in many applications of adaptive filtering. However, the computational complexity of this algorithm represents a major limitation in some applications that involve long filters, like echo cancellation. Moreover, the specific features of this application require good tracking capabilities and double-talk robustness for the adaptive algorithm, which further imply an optimization process on its parameters. In the case of most RLS-based algorithms, the performance can be controlled in terms of two main parameters, i.e., the forgetting factor and the regularization term. In this paper, we outline the influence of these parameters on the overall performance of the RLS algorithm and present several solutions to control their behavior, taking into account the specific requirements of echo cancellation application. The resulting variable forgetting factor RLS (VFF-RLS) and variable-regularized RLS (...
Circular differential microphone arrays (CDMAs) are characterized as compact superdirective beamf... more Circular differential microphone arrays (CDMAs) are characterized as compact superdirective beamformers whose beampatterns are almost frequency invariant. In contrast to linear differential microphone arrays (LDMAs) where the optimal steering direction is at the endfire, CDMAs provide almost perfect steering for all azimuthal directions. Herein, we present the design of a first-order CDMA in the time domain which is motivated by several aspects. First, time-domain implementation is important in some applications where minimal delay is required, such as real-time communications. Moreover, direct design in the time domain can reduce the computational efforts compared to the frequency-domain design, especially when short filters are sufficient. We present a design example for the time-domain first-order CDMA illustrating some of its fundamental properties as well as the equivalence to the frequency-domain alternative.
Sound spatial information benefits human listeners in reverberant environments. This paper deals ... more Sound spatial information benefits human listeners in reverberant environments. This paper deals with the problem of binaural dereverberation, which reduces reverberation and meanwhile preserves the sound spatial information at the binaural outputs. A widely linear (WL) filtering framework is adopted where the multiple real microphone signals are merged into complex signals. The desired binaural outputs are also converted into complex signals with one channel being its real part, and the other channel being its imaginary part. By doing so, we transform the problem of binaural dereverberation to one of monaural dereverberation. In such a framework, the complex late reverberation is modeled using the multichannel delayed WL prediction by fully taking advantage of the noncircularity of the complex signals. A maximum likelihood method is then developed to estimate the optimal prediction filter with the speech signal of interest being modeled by a complex normal distribution. The relationship between the proposed method and the weighted prediction error (WPE) method is also discussed. Finally, simulation results are provided to justify the effectiveness of the developed method.
The multilinear system framework allows for the exploitation of the system identification problem... more The multilinear system framework allows for the exploitation of the system identification problem from different perspectives in the context of various applications, such as nonlinear acoustic echo cancellation, multi-party audio conferencing, and video conferencing, in which the system could be modeled through parallel or cascaded filters. In this paper, we introduce different memoryless and memory structures that are described from a bilinear perspective. Following the memory structures, we develop the multilinear recursive least-squares algorithm by considering the Kronecker product decomposition concept. We have performed a set of simulations in the context of echo cancellation, aiming both long length impulse responses and the reverberation effect.
2018 16th International Workshop on Acoustic Signal Enhancement (IWAENC)
Differential microphone arrays (DMAs) have attracted great interest over the past two decades, si... more Differential microphone arrays (DMAs) have attracted great interest over the past two decades, since this type of arrays can form frequency-invariant beampatterns and achieve maximum directional gains with a given number of sensors. Generally, the design of DMA beamformers involves optimization of some performance measures such as the directivity factor (DF), front-to-back ratio (FBR), white noise gain (WNG), etc. In this paper, we develop approximate performance measures, which basically approximate the integral part in the exact performance measures with a weighted sum. This approximation gets finer and finer as more points are used. When applied to the design of DMAs, the major advantages of using these approximate measures is that the design problem is simplified and many differential beamformers, including commonly-used standard ones, can be easily derived.
Any good microphone array system requires a reliable beamforming algorithm at the outputs of the ... more Any good microphone array system requires a reliable beamforming algorithm at the outputs of the sensors to enhance a desired signal coming from a known direction. There are many ways to optimize the coefficients of this beamformer depending on what we want and the application at hand. Fundamentally, there are three large classes of conventional beamformers; they are the fixed, adaptive, and differential beamformers. In this chapter, we show how to derive most of their counterparts as well as new approaches with Kronecker product filters.
2016 IEEE International Workshop on Acoustic Signal Enhancement (IWAENC)
Differential microphone arrays (DMAs) have a great potential to overcome some of the problems of ... more Differential microphone arrays (DMAs) have a great potential to overcome some of the problems of additive arrays and provide high spatial gain relative to their small size. In this work, we present a time-domain formulation for implementing first-order DMAs, which is very important for some applications in which minimal delay is required, such as real-time communications. We present a design example for first-order DMAs illustrating some of the fundamental properties of the time-domain implementation as well as the equivalence to the frequency-domain implementation. Index Terms-Microphone arrays, differential microphone arrays (DMAs), time-domain broadband beamforming.
Separating independent speech sources from their convolutive mixtures in a reverberant acoustic e... more Separating independent speech sources from their convolutive mixtures in a reverberant acoustic environment is a challenging problem because of two difficulties: (a) very little is known about the source signals or the way they are mixed, and (b) both spatial interference from competing sources and temporal echoes due to room reverberation are observed in the mixtures. In this paper, after blindly identifying the acoustic MIMO system, we deal with spatial interference and temporal echoes in two different steps by converting an ¤ ¦ ¥ § MIMO system into ¤ SIMO systems. The performance is evaluated by simulations with measurements obtained in the varechoic chamber at Bell Labs.
Efficiently solving a system identification problem represents an important step in numerous impo... more Efficiently solving a system identification problem represents an important step in numerous important applications. In this framework, some of the most popular solutions rely on the Wiener filter, which is widely used in practice. Moreover, it also represents a benchmark for other related optimization problems. In this paper, new insights into the regularization of the Wiener filter are provided, which is a must in real-world scenarios. A proper regularization technique is of great importance, especially in challenging conditions, e.g., when operating in noisy environments and/or when only a low quantity of data is available for the estimation of the statistics. Different regularization methods are investigated in this paper, including several new solutions that fit very well for the identification of sparse and low-rank systems. Experimental results support the theoretical developments and indicate the efficiency of the proposed techniques.
Due to its fast convergence rate, the recursive least-squares (RLS) algorithm is very popular in ... more Due to its fast convergence rate, the recursive least-squares (RLS) algorithm is very popular in many applications of adaptive filtering. However, the computational complexity of this algorithm represents a major limitation in some applications that involve long filters, like echo cancellation. Moreover, the specific features of this application require good tracking capabilities and double-talk robustness for the adaptive algorithm, which further imply an optimization process on its parameters. In the case of most RLS-based algorithms, the performance can be controlled in terms of two main parameters, i.e., the forgetting factor and the regularization term. In this paper, we outline the influence of these parameters on the overall performance of the RLS algorithm and present several solutions to control their behavior, taking into account the specific requirements of echo cancellation application. The resulting variable forgetting factor RLS (VFF-RLS) and variable-regularized RLS (...
Circular differential microphone arrays (CDMAs) are characterized as compact superdirective beamf... more Circular differential microphone arrays (CDMAs) are characterized as compact superdirective beamformers whose beampatterns are almost frequency invariant. In contrast to linear differential microphone arrays (LDMAs) where the optimal steering direction is at the endfire, CDMAs provide almost perfect steering for all azimuthal directions. Herein, we present the design of a first-order CDMA in the time domain which is motivated by several aspects. First, time-domain implementation is important in some applications where minimal delay is required, such as real-time communications. Moreover, direct design in the time domain can reduce the computational efforts compared to the frequency-domain design, especially when short filters are sufficient. We present a design example for the time-domain first-order CDMA illustrating some of its fundamental properties as well as the equivalence to the frequency-domain alternative.
Sound spatial information benefits human listeners in reverberant environments. This paper deals ... more Sound spatial information benefits human listeners in reverberant environments. This paper deals with the problem of binaural dereverberation, which reduces reverberation and meanwhile preserves the sound spatial information at the binaural outputs. A widely linear (WL) filtering framework is adopted where the multiple real microphone signals are merged into complex signals. The desired binaural outputs are also converted into complex signals with one channel being its real part, and the other channel being its imaginary part. By doing so, we transform the problem of binaural dereverberation to one of monaural dereverberation. In such a framework, the complex late reverberation is modeled using the multichannel delayed WL prediction by fully taking advantage of the noncircularity of the complex signals. A maximum likelihood method is then developed to estimate the optimal prediction filter with the speech signal of interest being modeled by a complex normal distribution. The relationship between the proposed method and the weighted prediction error (WPE) method is also discussed. Finally, simulation results are provided to justify the effectiveness of the developed method.
The multilinear system framework allows for the exploitation of the system identification problem... more The multilinear system framework allows for the exploitation of the system identification problem from different perspectives in the context of various applications, such as nonlinear acoustic echo cancellation, multi-party audio conferencing, and video conferencing, in which the system could be modeled through parallel or cascaded filters. In this paper, we introduce different memoryless and memory structures that are described from a bilinear perspective. Following the memory structures, we develop the multilinear recursive least-squares algorithm by considering the Kronecker product decomposition concept. We have performed a set of simulations in the context of echo cancellation, aiming both long length impulse responses and the reverberation effect.
2018 16th International Workshop on Acoustic Signal Enhancement (IWAENC)
Differential microphone arrays (DMAs) have attracted great interest over the past two decades, si... more Differential microphone arrays (DMAs) have attracted great interest over the past two decades, since this type of arrays can form frequency-invariant beampatterns and achieve maximum directional gains with a given number of sensors. Generally, the design of DMA beamformers involves optimization of some performance measures such as the directivity factor (DF), front-to-back ratio (FBR), white noise gain (WNG), etc. In this paper, we develop approximate performance measures, which basically approximate the integral part in the exact performance measures with a weighted sum. This approximation gets finer and finer as more points are used. When applied to the design of DMAs, the major advantages of using these approximate measures is that the design problem is simplified and many differential beamformers, including commonly-used standard ones, can be easily derived.
Any good microphone array system requires a reliable beamforming algorithm at the outputs of the ... more Any good microphone array system requires a reliable beamforming algorithm at the outputs of the sensors to enhance a desired signal coming from a known direction. There are many ways to optimize the coefficients of this beamformer depending on what we want and the application at hand. Fundamentally, there are three large classes of conventional beamformers; they are the fixed, adaptive, and differential beamformers. In this chapter, we show how to derive most of their counterparts as well as new approaches with Kronecker product filters.
2016 IEEE International Workshop on Acoustic Signal Enhancement (IWAENC)
Differential microphone arrays (DMAs) have a great potential to overcome some of the problems of ... more Differential microphone arrays (DMAs) have a great potential to overcome some of the problems of additive arrays and provide high spatial gain relative to their small size. In this work, we present a time-domain formulation for implementing first-order DMAs, which is very important for some applications in which minimal delay is required, such as real-time communications. We present a design example for first-order DMAs illustrating some of the fundamental properties of the time-domain implementation as well as the equivalence to the frequency-domain implementation. Index Terms-Microphone arrays, differential microphone arrays (DMAs), time-domain broadband beamforming.
Separating independent speech sources from their convolutive mixtures in a reverberant acoustic e... more Separating independent speech sources from their convolutive mixtures in a reverberant acoustic environment is a challenging problem because of two difficulties: (a) very little is known about the source signals or the way they are mixed, and (b) both spatial interference from competing sources and temporal echoes due to room reverberation are observed in the mixtures. In this paper, after blindly identifying the acoustic MIMO system, we deal with spatial interference and temporal echoes in two different steps by converting an ¤ ¦ ¥ § MIMO system into ¤ SIMO systems. The performance is evaluated by simulations with measurements obtained in the varechoic chamber at Bell Labs.
Uploads
Papers by Jacob Benesty