Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032207
Ryosuke Takayama, Shuichi Arai
In recent years, multiresolutional time-frequency analysis has been widely studied in many fields. However, multiresolution analysis requires a large amount of calculations in accordance with the number of resolutions. Recently, the signal become higher resolution, so it tends to increase the amount of computation further. Gabor Transform (GT) is well-known for minimizing the product of time and frequency resolution, and is often used for that analysis. In this paper, we propose a method that can reduce the complexity cost of GT. First, we propose that the spectrum of high frequency resolution can be synthesized from spectra of high time resolution. Next, we show the method of determining the parameters based on the acceptable synthesis error. Finally, we show the effect of complexity cost reduction. The proposed method achieved lower calculation complexity down to 62.9% maximum.
{"title":"Fast multiresolution Gabor Transform based on synthesis of high frequency resolution spectrum from low frequency resolution spectra","authors":"Ryosuke Takayama, Shuichi Arai","doi":"10.1109/GlobalSIP.2014.7032207","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032207","url":null,"abstract":"In recent years, multiresolutional time-frequency analysis has been widely studied in many fields. However, multiresolution analysis requires a large amount of calculations in accordance with the number of resolutions. Recently, the signal become higher resolution, so it tends to increase the amount of computation further. Gabor Transform (GT) is well-known for minimizing the product of time and frequency resolution, and is often used for that analysis. In this paper, we propose a method that can reduce the complexity cost of GT. First, we propose that the spectrum of high frequency resolution can be synthesized from spectra of high time resolution. Next, we show the method of determining the parameters based on the acceptable synthesis error. Finally, we show the effect of complexity cost reduction. The proposed method achieved lower calculation complexity down to 62.9% maximum.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"112 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124763743","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032247
Dan Lv, A. Eslami, Shuguang Cui
Large-scale cascading failures could be triggered by very few initial failures, which lead to serious damages in complex networks. This paper presents a load-based cascade model to analyze the vulnerability of a complex system such as the power grid under a random single-node attack, where the Erdös-Renyi (ER) random graph is used as the network model since it captures the exponential degree distribution present in power grids. Based on a step-by-step analysis, we derive an estimate for the average failure ratio at each step, which provides a framework to track the impacts of cascading failures. Such a result is critical in designing robust and resilient complex networks.
{"title":"Load-based cascading failure analysis in finite Erdös-Rényi random networks","authors":"Dan Lv, A. Eslami, Shuguang Cui","doi":"10.1109/GlobalSIP.2014.7032247","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032247","url":null,"abstract":"Large-scale cascading failures could be triggered by very few initial failures, which lead to serious damages in complex networks. This paper presents a load-based cascade model to analyze the vulnerability of a complex system such as the power grid under a random single-node attack, where the Erdös-Renyi (ER) random graph is used as the network model since it captures the exponential degree distribution present in power grids. Based on a step-by-step analysis, we derive an estimate for the average failure ratio at each step, which provides a framework to track the impacts of cascading failures. Such a result is critical in designing robust and resilient complex networks.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123543455","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032163
A. Kazemipour, B. Babadi, Min Wu
In this paper, the performance of ℓ1-regularized Maximum-Likelihood estimator is investigated in sparse estimation of self-exciting processes. The underlying model includes a Generalized Linear Model (GLM) with Poisson observations and a parameter which is related to the covariates through a log-link. Kolmogorov-Smirnov and autocorrelation function tests are used as statistical goodness-of-fit measures. Results have shown a better performance of the regularized estimator both in the statistical sense and in the error norm. Application of the proposed algorithm to the LGN neuron firing data has successfully recovered the neurons' intrinsic frequencies.
{"title":"Sparse estimation of self-exciting point processes with application to LGN neural modeling","authors":"A. Kazemipour, B. Babadi, Min Wu","doi":"10.1109/GlobalSIP.2014.7032163","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032163","url":null,"abstract":"In this paper, the performance of ℓ1-regularized Maximum-Likelihood estimator is investigated in sparse estimation of self-exciting processes. The underlying model includes a Generalized Linear Model (GLM) with Poisson observations and a parameter which is related to the covariates through a log-link. Kolmogorov-Smirnov and autocorrelation function tests are used as statistical goodness-of-fit measures. Results have shown a better performance of the regularized estimator both in the statistical sense and in the error norm. Application of the proposed algorithm to the LGN neuron firing data has successfully recovered the neurons' intrinsic frequencies.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125507967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032306
Zhijin Qin, Yue Gao, Mark D. Plumbley, C. Parini, L. Cuthbert
Spectrum used for Machine-to-Machine (M2M) communications should be as cheap as possible or even free in order to connect billions of devices. Recently, both UK and US regulators have conducted trails and pilots to release the UHF TV spectrum for secondary licence-exempt applications. However, it is a very challenging task to implement wideband spectrum sensing in compact and low power M2M devices as high sampling rates are very expensive and difficult to achieve. In recent years, compressive sensing (CS) technique makes fast wideband spectrum sensing possible by taking samples at sub-Nyquist sampling rates. In this paper, we propose a two-step CS based spectrum sensing algorithm. In the first step, the CS is implemented in an SU and only part of the spectrum of interest is supposed to be sensed by an SU in each sensing period to reduce the complexity in the signal recovery process. In the second step, a denoising algorithm is proposed to improve the detection performance of spectrum sensing. The proposed two-step CS based spectrum sensing is compared with the traditional scheme and the theoretical curves.
{"title":"Efficient compressive spectrum sensing algorithm for M2M devices","authors":"Zhijin Qin, Yue Gao, Mark D. Plumbley, C. Parini, L. Cuthbert","doi":"10.1109/GlobalSIP.2014.7032306","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032306","url":null,"abstract":"Spectrum used for Machine-to-Machine (M2M) communications should be as cheap as possible or even free in order to connect billions of devices. Recently, both UK and US regulators have conducted trails and pilots to release the UHF TV spectrum for secondary licence-exempt applications. However, it is a very challenging task to implement wideband spectrum sensing in compact and low power M2M devices as high sampling rates are very expensive and difficult to achieve. In recent years, compressive sensing (CS) technique makes fast wideband spectrum sensing possible by taking samples at sub-Nyquist sampling rates. In this paper, we propose a two-step CS based spectrum sensing algorithm. In the first step, the CS is implemented in an SU and only part of the spectrum of interest is supposed to be sensed by an SU in each sensing period to reduce the complexity in the signal recovery process. In the second step, a denoising algorithm is proposed to improve the detection performance of spectrum sensing. The proposed two-step CS based spectrum sensing is compared with the traditional scheme and the theoretical curves.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130581425","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032219
S. Pequito, S. Kar, Antonio Pedro Aguiar
This paper studies the problem of identifying the minimum number of entities (agents), referred to as information gatherers, that are able to infer all the states in a dynamical social network. The information gatherers can be, for instance, service providers and the remaining agents the clients, each comprising several dynamic states associated with the services and personal information. The problem of identifying the minimum number of information gatherers can constitute a way to create coalitions to oversee the entire state of the system, and consequently the behavior of the agents in the social network. The dynamical social network is assumed to be modelled as a linear time-invariant system, and we will make use of the structural systems concept, i.e., by considering only the sparsity pattern (location of zeroes/non-zeroes) of the system coupling matrix. As a consequence, the design guarantees derived hold for almost all numerical parametric realizations of the system. In this paper, we show that this problem is NP-hard: in addition, we provide a reduction of the coalition problem to a minimum set covering problem that, in practice, leads to efficient (polynomial complexity) approximation schemes for solving the coalition problem with guaranteed optimality gaps. Finally, an example is provided which illustrates the analytical findings.
{"title":"Minimum number of information gatherers to ensure full observability of a dynamic social network: A structural systems approach","authors":"S. Pequito, S. Kar, Antonio Pedro Aguiar","doi":"10.1109/GlobalSIP.2014.7032219","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032219","url":null,"abstract":"This paper studies the problem of identifying the minimum number of entities (agents), referred to as information gatherers, that are able to infer all the states in a dynamical social network. The information gatherers can be, for instance, service providers and the remaining agents the clients, each comprising several dynamic states associated with the services and personal information. The problem of identifying the minimum number of information gatherers can constitute a way to create coalitions to oversee the entire state of the system, and consequently the behavior of the agents in the social network. The dynamical social network is assumed to be modelled as a linear time-invariant system, and we will make use of the structural systems concept, i.e., by considering only the sparsity pattern (location of zeroes/non-zeroes) of the system coupling matrix. As a consequence, the design guarantees derived hold for almost all numerical parametric realizations of the system. In this paper, we show that this problem is NP-hard: in addition, we provide a reduction of the coalition problem to a minimum set covering problem that, in practice, leads to efficient (polynomial complexity) approximation schemes for solving the coalition problem with guaranteed optimality gaps. Finally, an example is provided which illustrates the analytical findings.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"112 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121126120","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032330
G. Hellbourg, A. Chippendale, M. Kesteven, B. Jeffs
Interference mitigation is becoming necessary to make radio astronomy work in bands that are heavily used to support our modern lives. It is becoming particularly difficult to work at frequencies between 1100 MHz and 1300 MHz that are rapidly filling up with satellite navigation signals. Antenna array radio telescopes present the possibility of applying spatial Radio Frequency Interference (RFI) mitigation. Spatial filtering techniques for RFI mitigation have been introduced to radio astronomy in the last decades. The success of these techniques relies on accurately estimating the RFI spatial signature (or RFI subspace). The use of a reference antenna steering at the RFI sources provides a good estimation of the RFI subspace when correlated with an array radio telescope. However, predicting the evolution of this subspace with time is necessary in a multiple RFI scenario, when only a single RFI source can be monitored at a time with the reference antenna. This paper introduces a subspace tracking approach, based on the power method applied to covariance data. The RFI spatial signature estimates provided by the reference antenna are used to initialize the power method to support a faster convergence. Practical examples are shown, applying the method to real data from a single 188 element phased array feed designed for the Australian Square Kilometre Array Pathfinder (ASKAP) telescope.
{"title":"Reference antenna-based subspace tracking for RFI mitigation in radio astronomy","authors":"G. Hellbourg, A. Chippendale, M. Kesteven, B. Jeffs","doi":"10.1109/GlobalSIP.2014.7032330","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032330","url":null,"abstract":"Interference mitigation is becoming necessary to make radio astronomy work in bands that are heavily used to support our modern lives. It is becoming particularly difficult to work at frequencies between 1100 MHz and 1300 MHz that are rapidly filling up with satellite navigation signals. Antenna array radio telescopes present the possibility of applying spatial Radio Frequency Interference (RFI) mitigation. Spatial filtering techniques for RFI mitigation have been introduced to radio astronomy in the last decades. The success of these techniques relies on accurately estimating the RFI spatial signature (or RFI subspace). The use of a reference antenna steering at the RFI sources provides a good estimation of the RFI subspace when correlated with an array radio telescope. However, predicting the evolution of this subspace with time is necessary in a multiple RFI scenario, when only a single RFI source can be monitored at a time with the reference antenna. This paper introduces a subspace tracking approach, based on the power method applied to covariance data. The RFI spatial signature estimates provided by the reference antenna are used to initialize the power method to support a faster convergence. Practical examples are shown, applying the method to real data from a single 188 element phased array feed designed for the Australian Square Kilometre Array Pathfinder (ASKAP) telescope.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116481496","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032290
P. V. R. Ferreira, Rushabh Metha, A. Wyglinski
This paper proposes an adaptive modulation scheme using rain fading predictions obtained via Kalman filtering in order to mitigate the effects of rain on cognitive radio-based geostationary (GEO) satellites operating in the Ka-band. In the proposed scheme, the need for adaptation is identified prior to the rain attenuation event, allowing for enough time for the transmitter and receiver to reconfigure, which is a requirement when one of the communicating nodes are moving at a certain relative speed. We show that the bit error rate (BER) performance can be improved by two orders of magnitude for a system that accounts for the overall delay when adapting its modulation scheme based on the proposed predictor outputs.
{"title":"Cognitive radio-based geostationary satellite communications for Ka-band transmissions","authors":"P. V. R. Ferreira, Rushabh Metha, A. Wyglinski","doi":"10.1109/GlobalSIP.2014.7032290","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032290","url":null,"abstract":"This paper proposes an adaptive modulation scheme using rain fading predictions obtained via Kalman filtering in order to mitigate the effects of rain on cognitive radio-based geostationary (GEO) satellites operating in the Ka-band. In the proposed scheme, the need for adaptation is identified prior to the rain attenuation event, allowing for enough time for the transmitter and receiver to reconfigure, which is a requirement when one of the communicating nodes are moving at a certain relative speed. We show that the bit error rate (BER) performance can be improved by two orders of magnitude for a system that accounts for the overall delay when adapting its modulation scheme based on the proposed predictor outputs.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"58 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133835611","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032215
P. Bianchi, W. Hachem, F. Iutzeler
Consider a network where each agent has a private composite function (e.g. the sum of a smooth and a non-smooth function). The problem we address here is to And a minimize! of the aggregate cost (the sum of the agents functions) in a distributed manner. In this paper, we combine recent results on primal-dual optimization and coordinate descent to propose an asynchronous distributed algorithm for composite optimization.
{"title":"A stochastic primal-dual algorithm for distributed asynchronous composite optimization","authors":"P. Bianchi, W. Hachem, F. Iutzeler","doi":"10.1109/GlobalSIP.2014.7032215","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032215","url":null,"abstract":"Consider a network where each agent has a private composite function (e.g. the sum of a smooth and a non-smooth function). The problem we address here is to And a minimize! of the aggregate cost (the sum of the agents functions) in a distributed manner. In this paper, we combine recent results on primal-dual optimization and coordinate descent to propose an asynchronous distributed algorithm for composite optimization.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133621087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032226
E. Gelenbe
We consider a network composed of a finite set of communicating nodes that send individual particles to each other, and each particle can carry binary information. Though our main motivation is related to communications in nanonetworks with electrons that carry magnetic spin as the bipolar information, one can also imagine that the particles may be molecules that use chirality to convey information. Since it is difficult for a particle to carry an identifier that conveys the identity of the "source" or "destination", each node receives particles whose source cannot be ascertained since physical imperfections may result in particles being directed to the wrong destination in a manner that interferes with the correctly directed particles, and particles that should arrive at a node may be received by some other node. In addition we consider the effect of noise which randomly switches the polarity of particles, and in the case of magnetic spin we also have the effect of entanglement. We compute the probability of error in such a network, and estimate the flow of particles that is needed, and the average energy consumption per particle, to insure a correct reception of the binary data carried by the flow.
{"title":"Error and energy when communicating with spins","authors":"E. Gelenbe","doi":"10.1109/GlobalSIP.2014.7032226","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032226","url":null,"abstract":"We consider a network composed of a finite set of communicating nodes that send individual particles to each other, and each particle can carry binary information. Though our main motivation is related to communications in nanonetworks with electrons that carry magnetic spin as the bipolar information, one can also imagine that the particles may be molecules that use chirality to convey information. Since it is difficult for a particle to carry an identifier that conveys the identity of the \"source\" or \"destination\", each node receives particles whose source cannot be ascertained since physical imperfections may result in particles being directed to the wrong destination in a manner that interferes with the correctly directed particles, and particles that should arrive at a node may be received by some other node. In addition we consider the effect of noise which randomly switches the polarity of particles, and in the case of magnetic spin we also have the effect of entanglement. We compute the probability of error in such a network, and estimate the flow of particles that is needed, and the average energy consumption per particle, to insure a correct reception of the binary data carried by the flow.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"95 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132031746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2014-12-01DOI: 10.1109/GlobalSIP.2014.7032180
T. Higuchi, H. Kameoka
This paper proposes a novel method for simultaneously solving the problems of underdetermined blind source separation (BSS), source activity detection, dereverberation and direction-of-arrival (DOA) estimation by introducing an extension of the "multichannel factorial hidden Markov model (MFH-MM)." The MFHMM is an extension of the multichannel non-negative matrix factorization (NMF) modeL in which the basis spectra are allowed to vary over time according to the transitions of the hidden states. This model has allowed us to perform source separation, source activity detection and dereverberation in a unified manner. In our previous model, the spatial covariance of each source has been treated as a model parameter. This has led the entire generative model to have an unnecessarily high degree of freedom, and thus the parameter inference has been prone to getting trapped into undesired local optima. To reasonably restrict the solution space of the spatial covariance matrix of each source, we propose to describe it as a weighted sum of the fixed spatial covariance matrix corresponding to the discrete set of DOAs. Through the parameter inference, the proposed model allows us to simultaneously solve the problems of underdetermined BSS, source activity detection, dereverberation and DOA estimation. Experimental results revealed that the proposed method was superior to a previous method in terms of the signal-to-distortion ratios of separated signals.
{"title":"Unified approach for underdetermined BSS, VAD, dereverberation and DOA estimation with multichannel factorial HMM","authors":"T. Higuchi, H. Kameoka","doi":"10.1109/GlobalSIP.2014.7032180","DOIUrl":"https://doi.org/10.1109/GlobalSIP.2014.7032180","url":null,"abstract":"This paper proposes a novel method for simultaneously solving the problems of underdetermined blind source separation (BSS), source activity detection, dereverberation and direction-of-arrival (DOA) estimation by introducing an extension of the \"multichannel factorial hidden Markov model (MFH-MM).\" The MFHMM is an extension of the multichannel non-negative matrix factorization (NMF) modeL in which the basis spectra are allowed to vary over time according to the transitions of the hidden states. This model has allowed us to perform source separation, source activity detection and dereverberation in a unified manner. In our previous model, the spatial covariance of each source has been treated as a model parameter. This has led the entire generative model to have an unnecessarily high degree of freedom, and thus the parameter inference has been prone to getting trapped into undesired local optima. To reasonably restrict the solution space of the spatial covariance matrix of each source, we propose to describe it as a weighted sum of the fixed spatial covariance matrix corresponding to the discrete set of DOAs. Through the parameter inference, the proposed model allows us to simultaneously solve the problems of underdetermined BSS, source activity detection, dereverberation and DOA estimation. Experimental results revealed that the proposed method was superior to a previous method in terms of the signal-to-distortion ratios of separated signals.","PeriodicalId":362306,"journal":{"name":"2014 IEEE Global Conference on Signal and Information Processing (GlobalSIP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2014-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129443072","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：