Search

In this paper, we present audible (normal) speech and inaudible murmur hidden Markov models based automatic speech recognition using NAM microphone. The NAM (Non-Audible Murmur) microphone is a special device, which can be used for capturing inaudible murmur speech. The device is based on the stethoscope, which is used in medical science. By attaching the NAM microphone behind the talker's ear, we can receive very quietly uttered speech and perform automatic speech recognition in a conventional way. Privacy, robustness to environmental, and a useful tool for sound-impaired people noise belong to the advantages of the NAM microphone. Using adaptation techniques, we created hidden Markov models for inaudible speech and we performed automatic speech recognition. The achieved results are very promising, and prove the effectiveness of NAM microphone In this paper, we also introduce our work for recognizing normal speech using NAM microphone. The idea is to take advantage of noise robustness of NAM microphone. In our experiments, we achieved a 93.82% word accuracy in clean environment, and a 93.08% word accuracy in noisy environment. In this paper, we also introduce two techniques to intergrate inaudible murmur and audible speech recognition using NAM microphone. In both cases, we achieved a 92.01% word accuracy on average, which is a very promising result., EUSIPCO2004: the 12th European Signal Processing Conference, September 6-10, 2004, Vienna, Austria.

In this paper, we provide a new theoretical analysis of the amount of musical noise generated via generalized spectral subtraction based on higher-order statistics. Power spectral subtraction is the most commonly used spectral subtraction method, and in our previous study a musical noise assessment theory limited to the power spectral domain was proposed. Therefore, in this paper, we propose a generalization of our previous theory on spectral subtraction for arbitrary exponent parameters. We can thus compare the amount of musical noise between any exponent domains from the results of our analysis. We also clarify that less musical noise is generated when we choose the lower exponent spectral domain; this implies that there is no theoretical justification for using power/amplitude spectral subtraction., EUSIPCO2010: 18th European Signal Processing Conference, August 23-27, 2010, Aalborg, Denmark.

This paper proposes a on-line adaptive relaxation algorithm for inverse filter. The inverse filter of the conventional method often amplifies the background noise and line noise when the room transfer function changes. To resolve this problem, an iterative relaxation algorithm for inverse filter by truncated singular value decomposition (Adaptive TSVD) have been proposed. However，it is difficult to apply this relaxation method in a duration where the original signal contains the sound. Therefore, we extend Adaptive TSVD into an on-line type algorithm based on the observed signal at the only one control point，normalizing the observed signal with the original sound. The result of the simulation reveals that the proposed method can always carry out the relaxation process against to acoustic fluctuation in any time durations. Also, subjective evaluation in the real acoustic environment indicates that the sound quality improves without degrading the localization., EUSIPCO2004: the 12th European Signal Processing Conference, September 6-10, 2004, Vienna, Austria.

In this paper, blind separation and deconvolution (BSD) problem with binaural-sound mixtures is addressed. We have proposed two-stage blind separation and deconvolution algorithm, which consists of Single-Input Multiple-Output (SIMO)-model-based ICA (SIMO-ICA) and blind multichannel inverse filtering. In the previous report, we carried out simulations in the artificial mixing system and only showed that the proposed BSD can work theoretically. In order to evaluate the proposed method in more actual situations, we carried out BSD experiments assuming that speech sources are convolved with head related transfer functions (HRTFs). The simulation results reveal that the proposed BSD method can be effective in the separation and deconvolution even with binaural-sound mixtures., EUSIPCO2004: the 12th European Signal Processing Conference, September 6-10, 2004, Vienna, Austria.

In this paper, we consider the human/machine hands-free speech interface where the user voice is picked at a distance with a microphone array. It is assumed that the user is close enough to the machine to be considered as a point source. The noise is a diffuse background noise, created by all the sources present in the environment. The proposed method aims at suppressing the diffuse background noise efficiently without distorting the speech estimate. This method is a modification of a method combining frequency domain blind signal separation (FD-BSS) and Wiener filter based postprocessing. Contrary to the conventional approach, we modify the estimate of the diffuse background noise given by FDBSS before applying theWiener filter based post-processing. We also have to build a modified observation using the FDBSS in order to apply our modified post-processing. Simulation results show that the proposed approach can achieve a better speech enhancement, measured in term of word recognition in a speech recognition task, than the conventional Wiener filter based post-processing., EUSIPCO2009: the 17th European Signal Processing Conference, August 24-28, 2009, Glasgow, Scotland.

In this paper, we address the blind separation problem of binaural mixed signals, and propose a novel blind separation method using Single-Input-Multiple-Output-model-based independent component analysis (SIMO-ICA) with a self-generator (SG) for the initial filter. SIMO-ICA which has been proposed by the authors can separate mixed signals, not into monaural source signals but into SIMO-model-based signals from independent sources as they are at the microphones. Although this attractive feature of SIMO-ICA is beneficial to the binaural sound separation, SIMO-ICA has a serious drawback in its high sensitivity to the initial settings of the separation filter. In the proposed method, the SG functions as the preprocessor of SIMO-ICA, and it can provide a valid initial filter for SIMO-ICA. To evaluate its effectiveness, binaural sound separation experiments are carried out under a reverberant condition. The experimental results reveal that the separation performance of the proposed method is superior to those of conventional methods., EUSIPCO2005: the 13th European Signal Processing Conference, September 4-8, 2005, Antalya, Turkey.

Recent theoretical developments in the field of blind source separation (BSS) introduced the score function difference (SFD) and showed its close relationship with the differential of the mutual information. These results are of great interest for developing BSS methods since the mutual information can be seen as the canonical cost function for BSS. However these results were only proposed in the real case. In this paper, we extend these results to the complex case and propose a frequency domain approach of BSS based on the complex equivalent of the SFD., EUSIPCO2008: 16th European Signal Processing Conference, August 25-28, 2008, Lausanne, Switzerland.

The performance of automatic speech recognition for signals acquired through a hands-free speech interface is limited by the adverse effect of the noise and the reverberation. Frequency domain blind signal processing techniques, like blind signal separation, have been used with success for suppressing the noise in real situation but they usually do not take into account the reverberation. In this paper, we present a method based on frequency domain blind signal extraction that is aimed at suppressing both the adverse effect of the noise and the reverberation., EUSIPCO2010: 18th European Signal Processing Conference, August 23-27, 2010, Aalborg, Denmark.

This paper describes a new small-scale interface for a barge-in free spoken dialogue system combining a multichannel sound field control and a microphone array, in which the response sound from the system can be canceled out at the microphone points. The conventional method inhibits the user from moving because the system forces the user to stay in the fixed position where the response sound is reproduced. However, since the proposed method doesn't arrange the control points for the reproduction of the response sound to the user, the user's move is allowed. Furthermore, relaxation of the strict reproduction for the response sound enables us to design a stable system with fewer loudspeakers than the conventional method. Proposed method shows higher performances in the speech recognition experiments., EUSIPCO2005: the 13th European Signal Processing Conference, September 4-8, 2005, Antalya, Turkey.

Direction-of-arrival (DoA) estimation is a mature topic with decades of history. Despite the progress in the field, very few papers have looked at the problem of DoA estimation with unknown dynamic range. Consider the case of space exploration or near-field and far-field emitters. In such settings, the amplitude of the impinging wavefront can be much higher than the maximum recordable range of the sensor, resulting in information loss via clipping or sensor saturation. In this paper, we present a novel sensing approach for DoA estimation that exploits hardware-software co-design and is pivoted around the theme of unlimited sensing. On the hardware front, we capitalize on a radically new breed of analog-to-digital converters (ADCs) which, instead of saturating, produce modulo measurements. On the algorithmic front, we develop a mathematically guaranteed DoA estimation technique which is non-iterative and backwards compatible with existing DoA algorithms. Our computer experiments show the efficiency of our approach by estimating DoAs from signals which are orders of magnitude higher than the ADC threshold. Hence, our work paves a new path for inverse problems linked with DoA estimation and at the same time provides guidelines for new hardware development., Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Signal Processing Systems

Direction-of-arrival (DoA) estimation is a mature topic with decades of history. Despite the progress in the field, very few papers have looked at the problem of DoA estimation with unknown dynamic range. Consider the case of space exploration or near-field and far-field emitters. In such settings, the amplitude of the impinging wavefront can be much higher than the maximum recordable range of the sensor, resulting in information loss via clipping or sensor saturation. In this paper, we present a novel sensing approach for DoA estimation that exploits hardware-software co-design and is pivoted around the theme of unlimited sensing. On the hardware front, we capitalize on a radically new breed of analog-to-digital converters (ADCs) which, instead of saturating, produce modulo measurements. On the algorithmic front, we develop a mathematically guaranteed DoA estimation technique which is non-iterative and backwards compatible with existing DoA algorithms. Our computer experiments show the efficiency of our approach by estimating DoAs from signals which are orders of magnitude higher than the ADC threshold. Hence, our work paves a new path for inverse problems linked with DoA estimation and at the same time provides guidelines for new hardware development., Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Circuits and Systems

Atrial fibrillation (AF) is the most common arrhythmia in the heart. Two main types of AF are defined as paroxysmal and persistent. In this paper, we present a method to discriminate between the characteristics of paroxysmal and persistent using tensor decompositions of a multi-channel electrocardiogram (ECG) signal. For this purpose, ECG signals are segmented by applying a Hilbert transform on the thresholded signal. Dynamic time warping is used to align the separated segments of each channel and then a tensor is constructed with three dimensions as time, heartbeats and channels. A Canonical polyadic decomposition with rank 2 is computed from this tensor and the resulting loading vectors describe the characteristics of paroxysmal and persistent AF in these three dimensions. The time loading vector reveals the pattern of a single P wave or abnormal AF patterns. The heartbeat loading vector shows whether the pattern is present or absent in a specific beat. The results can be used to distinguish between the patterns in paroxysmal AF and persistent AF., Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Circuits and Systems

The use of wireless acoustic sensor networks (WASNs) has received increased attention over the last decade. The advantages of WASNs over stand-alone multi-microphone devices are that the microphone array is not anymore limited by the dimensions of a single device, and that microphones can be placed at arbitrary locations. One of the disadvantages, however, is that for many applications, like beamforming, the clocks of all devices in the network need to be synchronised and that the microphone gains need to be equalised. In this paper we will prove that a specific class of beamformers is clock-offset and gain mismatch invariant. The parameters for these beamformers (acoustic transfer function and power spectral density matrices) can be estimated directly from the uncalibrated microphone signals, instead of first synchronising the clocks and equalising the gains and then estimating them. The resulting beamformers are applied to the non-calibrated microphone signals. We will substantiate, by means of computer simulations, that the proposed approach gives identical results compared to the setup where microphone signals are first calibrated, so that clock-offset compensation and microphone gain equalisation becomes unnecessary., Circuits and Systems

Over the past decades, privacy-preservation has received considerable attention, not only as a consequence of regulations such as the General Data Protection Regulation in the EU, but also from the fact that people are more concerned about data abuse as the world is becoming increasingly digitized. In this paper we propose a convex optimization-based subspace perturbation approach to solve privacy-preserving distributed least squares problems. Based on the primal-dual method of multipliers, the introduced dual variables will only converge in a subspace determined by the graph topology and do not converge in its orthogonal complement. We, therefore, propose to exploit this property for privacy-preservation by using the nonconverging part of the dual variables to perturb the private data, thereby protecting it from being revealed. Moreover, we prove that the proposed approach is secure under both eavesdropping and passive adversaries. Computer simulations are conducted to demonstrate the benefits of the proposed approach through its convergence properties and accuracy., Circuits and Systems

Atrial electrograms are often used to gain understanding on the development of atrial fibrillation (AF). Using such electrograms, cardiologists can reconstruct how the depolarization wave-front propagates across the atrium. Knowing the exact moment at which the depolarization wavefront in the tissue reaches each electrode is an important aspect of such reconstruction. A common way to determine the LAT is based on the steepest deflection (SD) of the individual electrograms. However, the SD annotates each electrogram individually and is expected to be more prone to errors compared to approaches that would employ the data from the surrounding electrodes to estimate the LAT. As electrograms from neighboring electrodes tend to have rather similar morphology up to a delay, we propose in this paper to use the cross-correlation to find the pair-wise relative delays between electrograms. Instead of only using the direct neighbors we consider the array as a graph and involve higher order neighbors as well. Using a least-squares method, the absolute LATs can then be estimated from the calculated pair-wise relative delays. Simulated and clinically recorded electrograms are used to evaluate the proposed approach. From the simulated data it follows that the proposed approach outperforms the SD approach., Circuits and Systems

Atrial electrograms are often used to gain understanding on the development of atrial fibrillation (AF). Using such electrograms, cardiologists can reconstruct how the depolarization wave-front propagates across the atrium. Knowing the exact moment at which the depolarization wavefront in the tissue reaches each electrode is an important aspect of such reconstruction. A common way to determine the LAT is based on the steepest deflection (SD) of the individual electrograms. However, the SD annotates each electrogram individually and is expected to be more prone to errors compared to approaches that would employ the data from the surrounding electrodes to estimate the LAT. As electrograms from neighboring electrodes tend to have rather similar morphology up to a delay, we propose in this paper to use the cross-correlation to find the pair-wise relative delays between electrograms. Instead of only using the direct neighbors we consider the array as a graph and involve higher order neighbors as well. Using a least-squares method, the absolute LATs can then be estimated from the calculated pair-wise relative delays. Simulated and clinically recorded electrograms are used to evaluate the proposed approach. From the simulated data it follows that the proposed approach outperforms the SD approach., Multimedia Computing

Average consensus algorithms are used in many distributed systems such as distributed optimization, sensor fusion and the control of dynamic systems. Consensus algorithms converge through an explicit exchange of state variables. In some cases, however, the state variables are confidential. In this paper, a privacy-preserving asynchronous distributed average consensus method is proposed, which decomposes the initial values into two states; alpha states and beta states. These states are initialized such that their sum is twice the initial value. The alpha states are used to communicate with the other nodes, while the beta states are used internally. Although beta states are not shared, they are used in the update of the alpha states. Unlike differential privacy based methods, the proposed algorithm achieves the exact average consensus, while providing privacy to the initial values. Compared to the synchronous state decomposition algorithm, the convergence rate is improved without any privacy compromise. As the variances of coupling weights become infinitely large, the semi-honest adversary does not have any range to estimate the initial value of the nodes given that there is at least one coupling weight hidden from the adversary., Circuits and Systems

This paper aims at developing low-complexity solutions for super-resolution two-dimensional (2D) harmonic retrieval via covariance reconstruction. Given the collected sample covariance, a novel gridless compressed sensing approach is designed based on the atomic norm minimization (ANM) technique. The key is to perform a redundancy reduction (RR) transformation that effectively reduces the large problem size at hand, without loss of useful frequency information. For uncorrelated sources, the transformed 2D covariance matrices in the RR domain retain a salient structure, which permits a sparse representation over a matrix-form atom set with decoupled 1D frequency components. Accordingly, the decoupled ANM (DANM) framework can be applied for super-resolution 2D frequency estimation, at low computational complexity on the same order of the 1D case. An analysis of the complexity reduction of the proposed RR-D-ANM compared with benchmark methods is provided as well, which is verified by our simulation results, Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Signal Processing Systems

Atrial electrograms are often used to gain understanding on the development of atrial fibrillation (AF). Using such electrograms, cardiologists can reconstruct how the depolarization wave-front propagates across the atrium. Knowing the exact moment at which the depolarization wavefront in the tissue reaches each electrode is an important aspect of such reconstruction. A common way to determine the LAT is based on the steepest deflection (SD) of the individual electrograms. However, the SD annotates each electrogram individually and is expected to be more prone to errors compared to approaches that would employ the data from the surrounding electrodes to estimate the LAT. As electrograms from neighboring electrodes tend to have rather similar morphology up to a delay, we propose in this paper to use the cross-correlation to find the pair-wise relative delays between electrograms. Instead of only using the direct neighbors we consider the array as a graph and involve higher order neighbors as well. Using a least-squares method, the absolute LATs can then be estimated from the calculated pair-wise relative delays. Simulated and clinically recorded electrograms are used to evaluate the proposed approach. From the simulated data it follows that the proposed approach outperforms the SD approach., Multimedia Computing

Over the past decades, privacy-preservation has received considerable attention, not only as a consequence of regulations such as the General Data Protection Regulation in the EU, but also from the fact that people are more concerned about data abuse as the world is becoming increasingly digitized. In this paper we propose a convex optimization-based subspace perturbation approach to solve privacy-preserving distributed least squares problems. Based on the primal-dual method of multipliers, the introduced dual variables will only converge in a subspace determined by the graph topology and do not converge in its orthogonal complement. We, therefore, propose to exploit this property for privacy-preservation by using the nonconverging part of the dual variables to perturb the private data, thereby protecting it from being revealed. Moreover, we prove that the proposed approach is secure under both eavesdropping and passive adversaries. Computer simulations are conducted to demonstrate the benefits of the proposed approach through its convergence properties and accuracy., Circuits and Systems

Average consensus algorithms are used in many distributed systems such as distributed optimization, sensor fusion and the control of dynamic systems. Consensus algorithms converge through an explicit exchange of state variables. In some cases, however, the state variables are confidential. In this paper, a privacy-preserving asynchronous distributed average consensus method is proposed, which decomposes the initial values into two states; alpha states and beta states. These states are initialized such that their sum is twice the initial value. The alpha states are used to communicate with the other nodes, while the beta states are used internally. Although beta states are not shared, they are used in the update of the alpha states. Unlike differential privacy based methods, the proposed algorithm achieves the exact average consensus, while providing privacy to the initial values. Compared to the synchronous state decomposition algorithm, the convergence rate is improved without any privacy compromise. As the variances of coupling weights become infinitely large, the semi-honest adversary does not have any range to estimate the initial value of the nodes given that there is at least one coupling weight hidden from the adversary., Circuits and Systems

Atrial fibrillation (AF) is the most common arrhythmia in the heart. Two main types of AF are defined as paroxysmal and persistent. In this paper, we present a method to discriminate between the characteristics of paroxysmal and persistent using tensor decompositions of a multi-channel electrocardiogram (ECG) signal. For this purpose, ECG signals are segmented by applying a Hilbert transform on the thresholded signal. Dynamic time warping is used to align the separated segments of each channel and then a tensor is constructed with three dimensions as time, heartbeats and channels. A Canonical polyadic decomposition with rank 2 is computed from this tensor and the resulting loading vectors describe the characteristics of paroxysmal and persistent AF in these three dimensions. The time loading vector reveals the pattern of a single P wave or abnormal AF patterns. The heartbeat loading vector shows whether the pattern is present or absent in a specific beat. The results can be used to distinguish between the patterns in paroxysmal AF and persistent AF., Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Circuits and Systems

This paper aims at developing low-complexity solutions for super-resolution two-dimensional (2D) harmonic retrieval via covariance reconstruction. Given the collected sample covariance, a novel gridless compressed sensing approach is designed based on the atomic norm minimization (ANM) technique. The key is to perform a redundancy reduction (RR) transformation that effectively reduces the large problem size at hand, without loss of useful frequency information. For uncorrelated sources, the transformed 2D covariance matrices in the RR domain retain a salient structure, which permits a sparse representation over a matrix-form atom set with decoupled 1D frequency components. Accordingly, the decoupled ANM (DANM) framework can be applied for super-resolution 2D frequency estimation, at low computational complexity on the same order of the 1D case. An analysis of the complexity reduction of the proposed RR-D-ANM compared with benchmark methods is provided as well, which is verified by our simulation results, Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Circuits and Systems

Atrial electrograms are often used to gain understanding on the development of atrial fibrillation (AF). Using such electrograms, cardiologists can reconstruct how the depolarization wave-front propagates across the atrium. Knowing the exact moment at which the depolarization wavefront in the tissue reaches each electrode is an important aspect of such reconstruction. A common way to determine the LAT is based on the steepest deflection (SD) of the individual electrograms. However, the SD annotates each electrogram individually and is expected to be more prone to errors compared to approaches that would employ the data from the surrounding electrodes to estimate the LAT. As electrograms from neighboring electrodes tend to have rather similar morphology up to a delay, we propose in this paper to use the cross-correlation to find the pair-wise relative delays between electrograms. Instead of only using the direct neighbors we consider the array as a graph and involve higher order neighbors as well. Using a least-squares method, the absolute LATs can then be estimated from the calculated pair-wise relative delays. Simulated and clinically recorded electrograms are used to evaluate the proposed approach. From the simulated data it follows that the proposed approach outperforms the SD approach., Circuits and Systems

The use of wireless acoustic sensor networks (WASNs) has received increased attention over the last decade. The advantages of WASNs over stand-alone multi-microphone devices are that the microphone array is not anymore limited by the dimensions of a single device, and that microphones can be placed at arbitrary locations. One of the disadvantages, however, is that for many applications, like beamforming, the clocks of all devices in the network need to be synchronised and that the microphone gains need to be equalised. In this paper we will prove that a specific class of beamformers is clock-offset and gain mismatch invariant. The parameters for these beamformers (acoustic transfer function and power spectral density matrices) can be estimated directly from the uncalibrated microphone signals, instead of first synchronising the clocks and equalising the gains and then estimating them. The resulting beamformers are applied to the non-calibrated microphone signals. We will substantiate, by means of computer simulations, that the proposed approach gives identical results compared to the setup where microphone signals are first calibrated, so that clock-offset compensation and microphone gain equalisation becomes unnecessary., Circuits and Systems

In this paper, we address the singing voice separation problem and propose a novel unsupervised approach based on robust principal component analysis (RPCA) exploiting rank-1 constraint (CRPCA). RPCA is a recently proposed singing voice separation algorithm that can separate singing voice from the monaural recordings. Although RPCA has been successfully applied to singing voice separation task, it ignores the different characteristic values of singular value decomposition and computational complexity to minimize the nuclear norm for separating singing voice. Since rank-l constraint in the background music, as the background music has a large variation in richness than singing voice among different songs. Furthermore, rank-1 constraint can utilize a prior target rank to separate singing voice and background music from the mixture music signal. Accordingly, the proposed CRPCA method utilizes rank-1 constraint minimization of singular values in RPCA instead of minimizing the whole nuclear norm, which can not only describe the different values of singular value decomposition, but also the computation complexity is reduced. The experiment evaluation results reveal that CRPCA can achieve better separation performance than the previous methods, especially with regard to use time frequency masking on ccMixter and DSD100 datasets. In addition, the running time on CRPCA is shorter than others under the same conditions., identifier:https://dspace.jaist.ac.jp/dspace/handle/10119/15513

In this paper, we provide a new theoretical analysis of the amount of musical noise generated via generalized spectral subtraction based on higher-order statistics. Power spectral subtraction is the most commonly used spectral subtraction method, and in our previous study a musical noise assessment theory limited to the power spectral domain was proposed. Therefore, in this paper, we propose a generalization of our previous theory on spectral subtraction for arbitrary exponent parameters. We can thus compare the amount of musical noise between any exponent domains from the results of our analysis. We also clarify that less musical noise is generated when we choose the lower exponent spectral domain; this implies that there is no theoretical justification for using power/amplitude spectral subtraction.

In this paper, system-level performance of three different MMSE turbo MIMO equalization techniques is evaluated in realistic scenarios. Soft cancellation and minimum mean squared error filtering (SC/MMSE) turbo equalization and its complexity reduced version, turbo equalized diversity, is considered. Furthermore, another version of equalized diversity, turbo equalized diversitywith common SC/MMSE, which exploits the transmit diversity and coding gain through the cross-wise iterations over the decoding branches, is evaluated. The multi-dimensional channel soundingmeasurement data used for the simulations consists of snapshots measured in different channel conditions in terms of spatial and temporal properties. The system-level assessment is in terms of outage probabilities of the performance figures such as bit and frame error rates obtained by evaluating their cumulative probabilitydensities as well as throughput efficiencies using the field measurement data. It is found that the receivers considered in this paper can all provide reasonable system-level performance. However,turbo equalized diversity receiver is more sensitive to the channel conditions than the original SC/MMSE equalizer. It is also found that the performance gain, obtained from the cross-wise iteration over the decoding branches in the turbo equalized diversity with common SC/MMSE technique, is significant., 14th European Signal Processing Conference (EUSIPCO 2006), Florence, Italy, September 4-8, 2006, identifier:https://dspace.jaist.ac.jp/dspace/handle/10119/9116

In this paper, we present audible (normal) speech and inaudible murmur hidden Markov models based automatic speech recognition using NAM microphone. The NAM (Non-Audible Murmur) microphone is a special device, which can be used for capturing inaudible murmur speech. The device is based on the stethoscope, which is used in medical science. By attaching the NAM microphone behind the talker's ear, we can receive very quietly uttered speech and perform automatic speech recognition in a conventional way. Privacy, robustness to environmental, and a useful tool for sound-impaired people noise belong to the advantages of the NAM microphone. Using adaptation techniques, we created hidden Markov models for inaudible speech and we performed automatic speech recognition. The achieved results are very promising, and prove the effectiveness of NAM microphone In this paper, we also introduce our work for recognizing normal speech using NAM microphone. The idea is to take advantage of noise robustness of NAM microphone. In our experiments, we achieved a 93.82% word accuracy in clean environment, and a 93.08% word accuracy in noisy environment. In this paper, we also introduce two techniques to intergrate inaudible murmur and audible speech recognition using NAM microphone. In both cases, we achieved a 92.01% word accuracy on average, which is a very promising result.

Favorable propagation, dened as mutual orthogonality among the vector-valued channels to the terminals, is one of the key properties of the radio channel that is exploited in Massive MIMO. However, there has been little work that studies this topic in detail. In this paper, we first show that favorable propagation offers the most desirable scenario in terms of maximizing the sum-capacity. One useful proxy for whether propagation is favorable or not is the channel condition number. However, this proxy is not good for the case where the norms of the channel vectors are not equal. For this case, to evaluate how favorable the propagation oered by the channel is, we propose a distance from favorable propagation measure, which is the gap between the sum-capacity and the maximum capacity obtained under favorable propagation. Secondly, we examine how favorable the channels can be for two extreme scenarios: i.i.d. Rayleigh fading and uniform random line-of-sight(UR-LoS). Both environments offer (nearly) favorable propagation. Furthermore, to analyze the UR-LoS model, we propose an urns-and-balls model. This model is simple and explains the singular value spread characteristic of the UR-LoS model well.

Synchronization and localization are critical challenges for the coherent functioning of a wireless network of mobile nodes. In this paper, a novel joint non-linear range and affine time model is presented based on two way time stamp exchanges, extending an existing affine time-range model. For a pair of nodes, a closed form pairwise least squares solution is proposed for estimating pairwise range parameters, namely relative range, range rate and rate of range rate between the nodes, in addition to estimating the clock skews and the clock offsets. Extending these pair wise solutions to network wide ranging and clock synchronization, we present a central data fusion based global least squares solution. Furthermore, a new Constrained Cramer Rao Bound (CCRB) is derived for the joint time-range model and the proposed algorithms are shown to approach the theoretical limits asymptotically., Circuits and Systems, Electrical Engineering, Mathematics and Computer Science

Synchronization and localization are critical challenges for the coherent functioning of a wireless network of mobile nodes. In this paper, a novel joint non-linear range and affine time model is presented based on two way time stamp exchanges, extending an existing affine time-range model. For a pair of nodes, a closed form pairwise least squares solution is proposed for estimating pairwise range parameters, namely relative range, range rate and rate of range rate between the nodes, in addition to estimating the clock skews and the clock offsets. Extending these pair wise solutions to network wide ranging and clock synchronization, we present a central data fusion based global least squares solution. Furthermore, a new Constrained Cramer Rao Bound (CCRB) is derived for the joint time-range model and the proposed algorithms are shown to approach the theoretical limits asymptotically., Circuits and Systems, Electrical Engineering, Mathematics and Computer Science

In this paper delay-free distributed joint source-channel coding for communication of two correlated Gaussian sources over a Gaussian Multiple Access Channel (GMAC) are considered. Both discrete and hybrid discrete analog schemes arc proposed and optimized. The proposed schemes are noise robust and show promising performance which improve with increasing correlation., QC 20190925

The performance of automatic speech recognition for signals acquired through a hands-free speech interface is limited by the adverse effect of the noise and the reverberation. Frequency domain blind signal processing techniques, like blind signal separation, have been used with success for suppressing the noise in real situation but they usually do not take into account the reverberation. In this paper, we present a method based on frequency domain blind signal extraction that is aimed at suppressing both the adverse effect of the noise and the reverberation.

application/pdf, The performance of automatic speech recognition for signals acquired through a hands-free speech interface is limited by the adverse effect of the noise and the reverberation. Frequency domain blind signal processing techniques, like blind signal separation, have been used with success for suppressing the noise in real situation but they usually do not take into account the reverberation. In this paper, we present a method based on frequency domain blind signal extraction that is aimed at suppressing both the adverse effect of the noise and the reverberation., EUSIPCO2010: 18th European Signal Processing Conference, August 23-27, 2010, Aalborg, Denmark., conference paper

application/pdf, In this paper, we provide a new theoretical analysis of the amount of musical noise generated via generalized spectral subtraction based on higher-order statistics. Power spectral subtraction is the most commonly used spectral subtraction method, and in our previous study a musical noise assessment theory limited to the power spectral domain was proposed. Therefore, in this paper, we propose a generalization of our previous theory on spectral subtraction for arbitrary exponent parameters. We can thus compare the amount of musical noise between any exponent domains from the results of our analysis. We also clarify that less musical noise is generated when we choose the lower exponent spectral domain; this implies that there is no theoretical justification for using power/amplitude spectral subtraction., EUSIPCO2010: 18th European Signal Processing Conference, August 23-27, 2010, Aalborg, Denmark., conference paper

In this paper, we consider the human/machine hands-free speech interface where the user voice is picked at a distance with a microphone array. It is assumed that the user is close enough to the machine to be considered as a point source. The noise is a diffuse background noise, created by all the sources present in the environment. The proposed method aims at suppressing the diffuse background noise efficiently without distorting the speech estimate. This method is a modification of a method combining frequency domain blind signal separation (FD-BSS) and Wiener filter based postprocessing. Contrary to the conventional approach, we modify the estimate of the diffuse background noise given by FDBSS before applying theWiener filter based post-processing. We also have to build a modified observation using the FDBSS in order to apply our modified post-processing. Simulation results show that the proposed approach can achieve a better speech enhancement, measured in term of word recognition in a speech recognition task, than the conventional Wiener filter based post-processing.

application/pdf, In this paper, we consider the human/machine hands-free speech interface where the user voice is picked at a distance with a microphone array. It is assumed that the user is close enough to the machine to be considered as a point source. The noise is a diffuse background noise, created by all the sources present in the environment. The proposed method aims at suppressing the diffuse background noise efficiently without distorting the speech estimate. This method is a modification of a method combining frequency domain blind signal separation (FD-BSS) and Wiener filter based postprocessing. Contrary to the conventional approach, we modify the estimate of the diffuse background noise given by FDBSS before applying theWiener filter based post-processing. We also have to build a modified observation using the FDBSS in order to apply our modified post-processing. Simulation results show that the proposed approach can achieve a better speech enhancement, measured in term of word recognition in a speech recognition task, than the conventional Wiener filter based post-processing., EUSIPCO2009: the 17th European Signal Processing Conference, August 24-28, 2009, Glasgow, Scotland., conference paper

A primary objective of this paper is to examine the achievable rate region of frequency selective K-user multiple access channels (MACs) with soft-cancelling frequency domain minimum-mean-squared-error (SC FD-MMSE) turbo equalization.With the aid of the knowledge about the extrinsic information transfer (EXIT) characteristic of the equalizer, we study the problem of determining the optimal rate allocation (RAAL) to the K users, for which the sum rate is maximized, and introduce a heuristic approach based on the dynamic programming principle for solving this optimization problem. We also propose a practical code selection algorithm for the RAAL to theK users given the multidimensional EXIT functions of the equalizer. It is shown through simulations that with SC FD-MMSE turbo equalization, substantial throughputgain can be achieved with the proposed RAAL technique over automatic repeat request with fixed coding rate., International ITG Workshop on Smart Antennas - WSA 2009, February 16-18 2009, Berlin, Germany, identifier:https://dspace.jaist.ac.jp/dspace/handle/10119/9111

A novel arithmetic extended mapping technique is proposed in this paper for bit-interleaved coded modulation with iterative decoding (BICM-ID) with repetition codes as well as for their modifications. With the proposed mapping scheme, the computational complexity of BICM-ID is significantly reduced compared with conventional quadrature amplitude modulation (QAM) mapping schemes Performances of BICM-ID using the proposed extended mapping rules are also verified both by the extrinsic information transfer (EXIT) analysis and bit error rate (BER) simulations. The results show that with the proposed BICM-ID scheme, the code rate, at which, given the channel signal-to-noise power ratio (SNR) bit error rate can made arbitrarily small, is almost 90% of channel capacity, and such excellent property holds at relatively wide range of SNR values., International ITG Workshop on Smart Antennas - WSA 2009, February 16-18 2009, Berlin, Germany, identifier:https://dspace.jaist.ac.jp/dspace/handle/10119/9110

Many speech enhancement methods had been proposed to suppress the effects of noise or reverberation. Although most of methods aimed to enhance only noisy or reverberant speech, these cannot simultaneously enhance noisy reverberantspeech. This paper proposes a method for restoringthe power envelope from the noisy reverberant speech. This method is based on the MTF concept and does not require that the impulse response and noise conditions in the room acoustics (noisy reverberant environments) be measured. The proposed method suppresses the effects of reverberation and noise on the power envelopes by restoring the smeared MTF. We carried out 12,000 simulations of noisesuppression and dereverberation for noisy reverberant speech to objectively evaluate the proposed method. The results showed that the proposed method can simultaneously work well in both noise-suppression and dereverberation., identifier:https://dspace.jaist.ac.jp/dspace/handle/10119/9979

Recent theoretical developments in the field of blind source separation (BSS) introduced the score function difference (SFD) and showed its close relationship with the differential of the mutual information. These results are of great interest for developing BSS methods since the mutual information can be seen as the canonical cost function for BSS. However these results were only proposed in the real case. In this paper, we extend these results to the complex case and propose a frequency domain approach of BSS based on the complex equivalent of the SFD.

application/pdf, Recent theoretical developments in the field of blind source separation (BSS) introduced the score function difference (SFD) and showed its close relationship with the differential of the mutual information. These results are of great interest for developing BSS methods since the mutual information can be seen as the canonical cost function for BSS. However these results were only proposed in the real case. In this paper, we extend these results to the complex case and propose a frequency domain approach of BSS based on the complex equivalent of the SFD., EUSIPCO2008: 16th European Signal Processing Conference, August 25-28, 2008, Lausanne, Switzerland., conference paper

A goal of this paper is to combine the advantageous properties of multiple access codes (MAC) and space-time transmission (STT) in wireless relay communication systems. The use of MAC provides transmission with the separability of the signalstransmitted from multiple users, and STT the diversity gain. It is shown both by model-based numerical results and measurement data-based simulation results that the combined use of MAC and STT achieves significant improvement in biterror rate (BER) as well as signaling throughput of relay systems., International ITG/IEEE Workshop on Smart Antennas - WSA 2007, February 26-27 2007, Vienna, Austria, identifier:https://dspace.jaist.ac.jp/dspace/handle/10119/9115

In this paper, system-level performance of three different MMSE turbo MIMO equalization techniques is evaluated in realistic scenarios. Soft cancellation and minimum mean squared error filtering (SC/MMSE) turbo equalization and its complexity reduced version, turbo equalized diversity, is considered. Furthermore, another version of equalized diversity, turbo equalized diversitywith common SC/MMSE, which exploits the transmit diversity and coding gain through the cross-wise iterations over the decoding branches, is evaluated. The multi-dimensional channel soundingmeasurement data used for the simulations consists of snapshots measured in different channel conditions in terms of spatial and temporal properties. The system-level assessment is in terms of outage probabilities of the performance figures such as bit and frame error rates obtained by evaluating their cumulative probabilitydensities as well as throughput efficiencies using the field measurement data. It is found that the receivers considered in this paper can all provide reasonable system-level performance. However,turbo equalized diversity receiver is more sensitive to the channel conditions than the original SC/MMSE equalizer. It is also found that the performance gain, obtained from the cross-wise iteration over the decoding branches in the turbo equalized diversity with common SC/MMSE technique, is significant., 14th European Signal Processing Conference (EUSIPCO 2006), Florence, Italy, September 4-8, 2006

In this paper, we address the blind separation problem of binaural mixed signals, and propose a novel blind separation method using Single-Input-Multiple-Output-model-based independent component analysis (SIMO-ICA) with a self-generator (SG) for the initial filter. SIMO-ICA which has been proposed by the authors can separate mixed signals, not into monaural source signals but into SIMO-model-based signals from independent sources as they are at the microphones. Although this attractive feature of SIMO-ICA is beneficial to the binaural sound separation, SIMO-ICA has a serious drawback in its high sensitivity to the initial settings of the separation filter. In the proposed method, the SG functions as the preprocessor of SIMO-ICA, and it can provide a valid initial filter for SIMO-ICA. To evaluate its effectiveness, binaural sound separation experiments are carried out under a reverberant condition. The experimental results reveal that the separation performance of the proposed method is superior to those of conventional methods.

This paper describes a new small-scale interface for a barge-in free spoken dialogue system combining a multichannel sound field control and a microphone array, in which the response sound from the system can be canceled out at the microphone points. The conventional method inhibits the user from moving because the system forces the user to stay in the fixed position where the response sound is reproduced. However, since the proposed method doesn't arrange the control points for the reproduction of the response sound to the user, the user's move is allowed. Furthermore, relaxation of the strict reproduction for the response sound enables us to design a stable system with fewer loudspeakers than the conventional method. Proposed method shows higher performances in the speech recognition experiments.

application/pdf, This paper describes a new small-scale interface for a barge-in free spoken dialogue system combining a multichannel sound field control and a microphone array, in which the response sound from the system can be canceled out at the microphone points. The conventional method inhibits the user from moving because the system forces the user to stay in the fixed position where the response sound is reproduced. However, since the proposed method doesn't arrange the control points for the reproduction of the response sound to the user, the user's move is allowed. Furthermore, relaxation of the strict reproduction for the response sound enables us to design a stable system with fewer loudspeakers than the conventional method. Proposed method shows higher performances in the speech recognition experiments., EUSIPCO2005: the 13th European Signal Processing Conference, September 4-8, 2005, Antalya, Turkey., conference paper

application/pdf, In this paper, we address the blind separation problem of binaural mixed signals, and propose a novel blind separation method using Single-Input-Multiple-Output-model-based independent component analysis (SIMO-ICA) with a self-generator (SG) for the initial filter. SIMO-ICA which has been proposed by the authors can separate mixed signals, not into monaural source signals but into SIMO-model-based signals from independent sources as they are at the microphones. Although this attractive feature of SIMO-ICA is beneficial to the binaural sound separation, SIMO-ICA has a serious drawback in its high sensitivity to the initial settings of the separation filter. In the proposed method, the SG functions as the preprocessor of SIMO-ICA, and it can provide a valid initial filter for SIMO-ICA. To evaluate its effectiveness, binaural sound separation experiments are carried out under a reverberant condition. The experimental results reveal that the separation performance of the proposed method is superior to those of conventional methods., EUSIPCO2005: the 13th European Signal Processing Conference, September 4-8, 2005, Antalya, Turkey., conference paper

This paper proposes a on-line adaptive relaxation algorithm for inverse filter. The inverse filter of the conventional method often amplifies the background noise and line noise when the room transfer function changes. To resolve this problem, an iterative relaxation algorithm for inverse filter by truncated singular value decomposition (Adaptive TSVD) have been proposed. However，it is difficult to apply this relaxation method in a duration where the original signal contains the sound. Therefore, we extend Adaptive TSVD into an on-line type algorithm based on the observed signal at the only one control point，normalizing the observed signal with the original sound. The result of the simulation reveals that the proposed method can always carry out the relaxation process against to acoustic fluctuation in any time durations. Also, subjective evaluation in the real acoustic environment indicates that the sound quality improves without degrading the localization.