The Experts below are selected from a list of 294 Experts worldwide ranked by ideXlab platform
Y. Yamamoto - One of the best experts on this subject based on the ideXlab platform.
-
h optimal approximation for Causal spline interpolation
Signal Processing, 2011Co-Authors: M. Nagahara, Y. YamamotoAbstract:In this paper, we give a Causal solution to the problem of spline interpolation using H^~ optimal approximation. Generally speaking, spline interpolation requires filtering the whole sampled data, the past and the future, to reconstruct the inter-sample values. This leads to non-Causality of the filter, and this becomes a critical issue for real-time applications. Our objective here is to derive a Causal System which approximates spline interpolation by H^~ optimization for the filter. The advantage of H^~ optimization is that it can address uncertainty in the input signals to be interpolated in design, and hence the optimized System has robustness property against signal uncertainty. We give a closed-form solution to the H^~ optimization in the case of the cubic splines. For higher-order splines, the optimal filter can be effectively solved by a numerical computation. We also show that the optimal FIR (finite impulse response) filter can be designed by an LMI (linear matrix inequality), which can also be effectively solved numerically. A design example is presented to illustrate the result.
-
Causal Spline Interpolation by H# Optimization
2007 IEEE International Conference on Acoustics Speech and Signal Processing - ICASSP '07, 2007Co-Authors: M. Nagahara, Y. YamamotoAbstract:Spline interpolation Systems generally contain non-Causal filters, and hence such Systems are difficult to use for real-time processing. Our objective is to design a Causal System which approximates spline interpolation. This is formulated as a problem of designing a stable inverse of a System with unstable zeros. For this purpose, we adopt H00 optimization. We give a closed form solution to the H∞ optimization in the case of the cubic spline. For higher order splines, the optimal filter can be effectively solved by a numerical computation. We also show that the optimal FIR (finite impulse response) filter can be designed by an LMI (linear matrix inequality), which can also be effectively solved numerically. A design example is presented to illustrate the result.
-
Optimal wavelet expansion via sampled-data H∞ control theory
SICE Annual Conference 2007, 2007Co-Authors: M. Ogura, M. Nagahara, Y. YamamotoAbstract:Wavelet expansion of an L2 signal requires the L2 inner product of the original signal and a scaling function. In digital signal processing, it is common to use sampled data of continuous-time signals instead of the inner product. This however causes a large reconstruction error, called "wavelet crime." We therefore design a Causal System which produces an approximation of the inner product via sampled-data H∞ control theory. We then make extensions to a multi-rate and a multi-wavelet case. By numerical examples, we show the effectiveness of the proposed method.
-
Causal Spline Interpolation by Hα Optimization
2006 SICE-ICASE International Joint Conference, 2006Co-Authors: M. Nagahara, T. Wada, Y. YamamotoAbstract:Spline interpolation Systems generally contain non-Causal filters, and hence it is difficult to use such Systems for real-time processing. Our objective is to design a Causal System which approximates spline interpolation. This is formulated as a problem of designing a stable inverse of a System with unstable zeros. For this purpose, we adopt H alpha optimization. The Halpha-optimal inverse System can be effectively solved by standard MATLAB routines, and hence Causal spline interpolation is obtained. A numerical example is presented to illustrate the result
Deliang Wang - One of the best experts on this subject based on the ideXlab platform.
-
Learning Complex Spectral Mapping With Gated Convolutional Recurrent Networks for Monaural Speech Enhancement
IEEE ACM Transactions on Audio Speech and Language Processing, 2020Co-Authors: Deliang WangAbstract:Phase is important for perceptual quality of speech. However, it seems intractable to directly estimate phase spectra through supervised learning due to their lack of spectrotemporal structure in it. Complex spectral mapping aims to estimate the real and imaginary spectrograms of clean speech from those of noisy speech, which simultaneously enhances magnitude and phase responses of speech. Inspired by multi-task learning, we propose a gated convolutional recurrent network (GCRN) for complex spectral mapping, which amounts to a Causal System for monaural speech enhancement. Our experimental results suggest that the proposed GCRN substantially outperforms an existing convolutional neural network (CNN) for complex spectral mapping in terms of both objective speech intelligibility and quality. Moreover, the proposed approach yields significantly higher STOI and PESQ than magnitude spectral mapping and complex ratio masking. We also find that complex spectral mapping with the proposed GCRN provides an effective phase estimate.
-
Causal Deep CASA for Monaural Talker-Independent Speaker Separation
IEEE ACM Transactions on Audio Speech and Language Processing, 2020Co-Authors: Deliang WangAbstract:Talker-independent monaural speaker separation aims to separate concurrent speakers from a single-microphone recording. Inspired by human auditory scene analysis (ASA) mechanisms, a two-stage deep CASA approach has been proposed recently to address this problem, which achieves state-of-the-art results in separating mixtures of two or three speakers. A main limitation of deep CASA is that it is a non-Causal System, while many speech processing applications, e.g., telecommunication and hearing prosthesis, require Causal processing. In this study, we propose a Causal version of deep CASA to address this limitation. First, we modify temporal connections, normalization and clustering algorithms in deep CASA so that no future information is used throughout the deep network. We then train a $C$-speaker ($C \geq 2$) deep CASA System in a speaker-number-independent fashion, generalizable to speech mixtures with up to $C$ speakers without the prior knowledge about the speaker number. Experimental results show that Causal deep CASA achieves excellent speaker separation performance with known or unknown speaker numbers.
-
Real-time Speech Enhancement Using an Efficient Convolutional Recurrent Network for Dual-microphone Mobile Phones in Close-talk Scenarios
ICASSP 2019 - 2019 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2019Co-Authors: Xueliang Zhang, Deliang WangAbstract:In mobile speech communication, the quality and intelligibility of the received speech can be severely degraded by background noise if the far-end talker is in an adverse acoustic environment. Therefore, speech enhancement algorithms are typically integrated into mobile phones to remove background noise. In this paper, we propose a novel deep learning based framework for real-time speech enhancement on dual-microphone mobile phones in a close-talk scenario. It incorporates a convolutional recurrent network (CRN) with high computational efficiency. In addition, the framework amounts to a Causal System, which is necessary for real-time processing on mobile phones. We find that the proposed approach consistently outperforms a deep neural network (DNN) based method, as well as two traditional methods for speech enhancement.
-
Complex Spectral Mapping with a Convolutional Recurrent Network for Monaural Speech Enhancement
ICASSP 2019 - 2019 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2019Co-Authors: Deliang WangAbstract:Phase is important for perceptual quality in speech enhancement. However, it seems intractable to directly estimate phase spectrogram through supervised learning due to lack of clear structure in phase spectrogram. Complex spectral mapping aims to estimate the real and imaginary spectrograms of clean speech from those of noisy speech, which simultaneously enhances magnitude and phase responses of noisy speech. In this paper, we propose a new convolutional recurrent network (CRN) for complex spectral mapping, which leads to a Causal System for noise- and speaker-independent speech enhancement. In terms of objective intelligibility and perceptual quality, the proposed CRN significantly outperforms an existing convolutional neural network (CNN) for complex spectral mapping, as well as a strong CRN for magnitude spectral mapping. We additionally incorporate a newly-developed group strategy to substantially reduce the number of trainable parameters and the computational cost without sacrificing performance.
-
Late Reverberation Suppression Using Recurrent Neural Networks with Long Short-Term Memory
2018 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2018Co-Authors: Yan Zhao, Deliang Wang, Buye Xu, Tao ZhangAbstract:Human speech is usually distorted by room reverberation. These corruptions degrade speech quality and intelligibility, especially under a long reverberation time, and they also pose a serious problem for many speech-related applications such as automatic speech recognition. In this paper, we propose a supervised speech dereverberation algorithm that models late reverberation using a recurrent neural network (RNN) with long short-term memory (LSTM). By taking advantage of LSTM's ability to capture a long history, late reverberation can be effectively removed by the proposed approach. Systematic evaluations indicate that our approach improves the quality of reverberant speech in a wide range of reverberant conditions. Moreover, the proposed System is a Causal System, which can be applied in real-time applications.
M. Nagahara - One of the best experts on this subject based on the ideXlab platform.
-
h optimal approximation for Causal spline interpolation
Signal Processing, 2011Co-Authors: M. Nagahara, Y. YamamotoAbstract:In this paper, we give a Causal solution to the problem of spline interpolation using H^~ optimal approximation. Generally speaking, spline interpolation requires filtering the whole sampled data, the past and the future, to reconstruct the inter-sample values. This leads to non-Causality of the filter, and this becomes a critical issue for real-time applications. Our objective here is to derive a Causal System which approximates spline interpolation by H^~ optimization for the filter. The advantage of H^~ optimization is that it can address uncertainty in the input signals to be interpolated in design, and hence the optimized System has robustness property against signal uncertainty. We give a closed-form solution to the H^~ optimization in the case of the cubic splines. For higher-order splines, the optimal filter can be effectively solved by a numerical computation. We also show that the optimal FIR (finite impulse response) filter can be designed by an LMI (linear matrix inequality), which can also be effectively solved numerically. A design example is presented to illustrate the result.
-
Causal Spline Interpolation by H# Optimization
2007 IEEE International Conference on Acoustics Speech and Signal Processing - ICASSP '07, 2007Co-Authors: M. Nagahara, Y. YamamotoAbstract:Spline interpolation Systems generally contain non-Causal filters, and hence such Systems are difficult to use for real-time processing. Our objective is to design a Causal System which approximates spline interpolation. This is formulated as a problem of designing a stable inverse of a System with unstable zeros. For this purpose, we adopt H00 optimization. We give a closed form solution to the H∞ optimization in the case of the cubic spline. For higher order splines, the optimal filter can be effectively solved by a numerical computation. We also show that the optimal FIR (finite impulse response) filter can be designed by an LMI (linear matrix inequality), which can also be effectively solved numerically. A design example is presented to illustrate the result.
-
Optimal wavelet expansion via sampled-data H∞ control theory
SICE Annual Conference 2007, 2007Co-Authors: M. Ogura, M. Nagahara, Y. YamamotoAbstract:Wavelet expansion of an L2 signal requires the L2 inner product of the original signal and a scaling function. In digital signal processing, it is common to use sampled data of continuous-time signals instead of the inner product. This however causes a large reconstruction error, called "wavelet crime." We therefore design a Causal System which produces an approximation of the inner product via sampled-data H∞ control theory. We then make extensions to a multi-rate and a multi-wavelet case. By numerical examples, we show the effectiveness of the proposed method.
-
Causal Spline Interpolation by Hα Optimization
2006 SICE-ICASE International Joint Conference, 2006Co-Authors: M. Nagahara, T. Wada, Y. YamamotoAbstract:Spline interpolation Systems generally contain non-Causal filters, and hence it is difficult to use such Systems for real-time processing. Our objective is to design a Causal System which approximates spline interpolation. This is formulated as a problem of designing a stable inverse of a System with unstable zeros. For this purpose, we adopt H alpha optimization. The Halpha-optimal inverse System can be effectively solved by standard MATLAB routines, and hence Causal spline interpolation is obtained. A numerical example is presented to illustrate the result
Y. Jianchao - One of the best experts on this subject based on the ideXlab platform.
-
Alignment of non-texture video frames using Kalman filter
IET Computer Vision, 2011Co-Authors: Y. JianchaoAbstract:Video registration in the presence of several segments of non-texture frames remains a challenging issue, although a wide repertoire of image registration algorithms was developed over the last two decades. In this study, the authors proposed an effective Causal scheme of video frames registration, and implemented it by the modified two frames intensity match algorithm. In order to achieve its real-time performance, the authors avoid handling a much more complex appearance model in this Causal System by using registration parameters as measurement values, which are estimated directly from two consecutive frames via intensity matching. The simple linear and Gaussian System is thus derived and an efficient Kalman filter algorithm can be utilised. As the Kalman filter naturally incorporated the temporal information contained in a video into the estimation of registration parameters, the algorithm developed in this study is quite robust and has good performance even if the processed video contains several segments of non-texture frames. © 2011 The Institution of Engineering and Technology.
-
Alignment of non-texture video frames using kalman filter
IET Computer Vision, 2011Co-Authors: Y. JianchaoAbstract:Video registration in the presence of several segments of non-texture frames remains a challenging issue, although a wide repertoire of image registration algorithms was developed over the last two decades. In this study, the authors proposed an effective Causal scheme of video frames registration, and implemented it by the modified two frames intensity match algorithm. In order to achieve its real-time performance, the authors avoid handling a much more complex appearance model in this Causal System by using registration parameters as measurement values, which are estimated directly from two consecutive frames via intensity matching. The simple linear and Gaussian System is thus derived and an efficient Kalman filter algorithm can be utilised. As the Kalman filter naturally incorporated the temporal information contained in a video into the estimation of registration parameters, the algorithm developed in this study is quite robust and has good performance even if the processed video contains several segments of non-texture frames.
Daniel Mayer - One of the best experts on this subject based on the ideXlab platform.
-
State Energy-Based Approach as a Tool for Design and Simulation of Linear and Nonlinear Systems
Analysis Control and Optimal Operations in Hybrid Power Systems, 2013Co-Authors: Milan Stork, Josef Hrusak, Daniel MayerAbstract:This chapter deals with a new problem of physical correctness detection in the area of strictly Causal System representations. The starting point is energy and the assumption that a System can be represented by a proper interconnection. The interconnection or, better, the interaction between physical Systems can be described in terms of power exchange through power ports. The proposed approach to the problem solution is based on generalization of Tellegen’s theorem well known from electrical engineering. Consequently, mathematically as well as physically correct results are obtained. The contribution is mainly concerned with presentation of a new structural approach to analysis and synthesis of linear and nonlinear Causal Systems. It has been proven that complete analysis of System behavior reduces to two independent tests: the monotonicity test of abstract state space energy and that of complete state observability, eventually of its dual, i.e., complete state controllability property. For comparison, the example of port-Hamiltonian approach is also presented.
-
Non-degenerate dissipative structures transformation and simulation
2013 International Conference on Applied Electronics, 2013Co-Authors: Milan Stork, Josef Hrusak, Daniel MayerAbstract:Computation complexity of a broad variety of practical design problems is known to be strongly depending on an algebraic complexity of corresponding mathematical System representations. Especially some vector-matrix models are frequently used in numerous interdisciplinary fields. One way to overcome the complexity problems is based on some special algebraic structures of low order model approximations, such as e.g. balanced representations. Another approach based on the concept of sparse matrices has also become very popular. As a very successful special case of sparse matrix based approach a class of tridiagonal System representations [1] has found applications in solution of partial differential equations, digital signal processing, image processing, computational fluid dynamics, spline curve fitting and many others. In this contribution a generalized sparse matrix motivated multi-diagonal method is proposed and some new results, based on state space energy motivated Causal System representations are presented, too [2].
-
Physical correctness of System representations based on generalized Tellegen principle
2009 16th International Conference on Digital Signal Processing, 2009Co-Authors: Josef Hrusak, Milan Stork, Daniel MayerAbstract:The paper deals with a new problem of physical correctness detection in the area of strictly Causal System representations. The proposed approach to the problem solution is based on generalization of Tellegen's theorem well known from electrical engineering. Consequently, mathematically as well as physically correct results are obtained. The contribution is mainly concerned with presentation of a new structural approach to analysis and synthesis of linear and non-linear Causal Systems. It has been proven that complete analysis of instability, conservativity, dissipativity, anti-dissipativity, stability, asymptotic stability and chaoticity reduces to two independent tests: the monotonicity test of abstract state space energy and that of complete state observability, evtl. of its dual, i.e. complete state controllability property.