Pub Date : 2019-09-01DOI: 10.1109/ICASSP.2019.8682403
Amit Das, Jinyu Li, Changliang Liu, Y. Gong
Acoustic models are domain dependent and do not perform well if there is a mismatch between training and test conditions. As an alternative, the Mixture of Experts (MoE) model was introduced for multi-domain modeling. It combines the outputs of several domain specific models (or experts) using a gating network. However, one drawback is that the gating network directly uses raw features and is unaware of the state of the experts. In this work, we propose several alternatives to improve the MoE model. First, to make our MoE model state-aware, we use outputs of experts as inputs to the gating network. Then we show that vector based interpolation of the mixture weights is more effective than scalar interpolation. Second, we show that directly learning the mixture weights without using any complex gating is still effective. Finally, we introduce a hybrid attention model that uses the logits and mixture weights from the previous time step to generate the mixture weights at the current time. Our best proposed model outperforms a baseline model using LSTM based gating achieving about 20.48% relative reduction in word error rate (WER). Moreover, it beats an oracle model which picks the best expert for a given test condition.
{"title":"Universal Acoustic Modeling Using Neural Mixture Models","authors":"Amit Das, Jinyu Li, Changliang Liu, Y. Gong","doi":"10.1109/ICASSP.2019.8682403","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682403","url":null,"abstract":"Acoustic models are domain dependent and do not perform well if there is a mismatch between training and test conditions. As an alternative, the Mixture of Experts (MoE) model was introduced for multi-domain modeling. It combines the outputs of several domain specific models (or experts) using a gating network. However, one drawback is that the gating network directly uses raw features and is unaware of the state of the experts. In this work, we propose several alternatives to improve the MoE model. First, to make our MoE model state-aware, we use outputs of experts as inputs to the gating network. Then we show that vector based interpolation of the mixture weights is more effective than scalar interpolation. Second, we show that directly learning the mixture weights without using any complex gating is still effective. Finally, we introduce a hybrid attention model that uses the logits and mixture weights from the previous time step to generate the mixture weights at the current time. Our best proposed model outperforms a baseline model using LSTM based gating achieving about 20.48% relative reduction in word error rate (WER). Moreover, it beats an oracle model which picks the best expert for a given test condition.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"30 1","pages":"5681-5685"},"PeriodicalIF":0.0,"publicationDate":"2019-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88462698","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 : 2019-06-06DOI: 10.1109/ICASSP.2019.8682916
Zhaocheng Huang, J. Epps, Dale Joachim
Detection of depression from speech has attracted significant research attention in recent years but remains a challenge, particularly for speech from diverse smartphones in natural environments. This paper proposes two sets of novel features based on speech landmark bigrams associated with abrupt speech articulatory events for depression detection from smartphone audio recordings. Combined with techniques adapted from natural language text processing, the proposed features further exploit landmark bigrams by discovering latent articulatory events. Experimental results on a large, naturalistic corpus containing various spoken tasks recorded from diverse smartphones suggest that speech landmark bigram features provide a 30.1% relative improvement in F1 (depressed) relative to an acoustic feature baseline system. As might be expected, a key finding was the importance of tailoring the choice of landmark bigrams to each elicitation task, revealing that different aspects of speech articulation are elicited by different tasks, which can be effectively captured by the landmark approaches.
{"title":"Speech Landmark Bigrams for Depression Detection from Naturalistic Smartphone Speech","authors":"Zhaocheng Huang, J. Epps, Dale Joachim","doi":"10.1109/ICASSP.2019.8682916","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682916","url":null,"abstract":"Detection of depression from speech has attracted significant research attention in recent years but remains a challenge, particularly for speech from diverse smartphones in natural environments. This paper proposes two sets of novel features based on speech landmark bigrams associated with abrupt speech articulatory events for depression detection from smartphone audio recordings. Combined with techniques adapted from natural language text processing, the proposed features further exploit landmark bigrams by discovering latent articulatory events. Experimental results on a large, naturalistic corpus containing various spoken tasks recorded from diverse smartphones suggest that speech landmark bigram features provide a 30.1% relative improvement in F1 (depressed) relative to an acoustic feature baseline system. As might be expected, a key finding was the importance of tailoring the choice of landmark bigrams to each elicitation task, revealing that different aspects of speech articulation are elicited by different tasks, which can be effectively captured by the landmark approaches.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"3 10 1","pages":"5856-5860"},"PeriodicalIF":0.0,"publicationDate":"2019-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81377779","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 : 2019-05-27DOI: 10.1109/ICASSP.2019.8682657
Michael Muma, W. Zeng, A. Zoubir
Conventional approaches to matrix completion are sensitive to outliers and impulsive noise. This paper develops robust and computationally efficient M-estimation based matrix completion algorithms. By appropriately arranging the observed entries, and then applying alternating minimization, the robust matrix completion problem is converted into a set of regression M-estimation problems. Making use of differentiable loss functions, the proposed algorithm overcomes a weakness of the ℓp-loss (p ≤ 1), which easily gets stuck in an inferior point. We prove that our algorithm converges to a stationary point of the nonconvex problem. Huber’s joint M-estimate of regression and scale can be used as a robust starting point for Tukey’s redescending M-estimator of regression based on an auxiliary scale. Numerical experiments on synthetic and real-world data demonstrate the superiority to state-of-the-art approaches.
{"title":"Robust M-estimation Based Matrix Completion","authors":"Michael Muma, W. Zeng, A. Zoubir","doi":"10.1109/ICASSP.2019.8682657","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682657","url":null,"abstract":"Conventional approaches to matrix completion are sensitive to outliers and impulsive noise. This paper develops robust and computationally efficient M-estimation based matrix completion algorithms. By appropriately arranging the observed entries, and then applying alternating minimization, the robust matrix completion problem is converted into a set of regression M-estimation problems. Making use of differentiable loss functions, the proposed algorithm overcomes a weakness of the ℓp-loss (p ≤ 1), which easily gets stuck in an inferior point. We prove that our algorithm converges to a stationary point of the nonconvex problem. Huber’s joint M-estimate of regression and scale can be used as a robust starting point for Tukey’s redescending M-estimator of regression based on an auxiliary scale. Numerical experiments on synthetic and real-world data demonstrate the superiority to state-of-the-art approaches.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"31 1","pages":"5476-5480"},"PeriodicalIF":0.0,"publicationDate":"2019-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81231635","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 : 2019-05-27DOI: 10.1109/ICASSP.2019.8682680
A. V. Singh, K. Chaudhury
Consider a network with N nodes in d dimensions, and M overlapping subsets P1, ⋯,PM (subnetworks). Assume that the nodes in a given Pi are observed in a local coordinate system. We wish to register the subnetworks using the knowledge of the observed coordinates. More precisely, we want to compute the positions of the N nodes in a global coordinate system, given P1, ⋯, PM and the corresponding local coordinates. Among other applications, this problem arises in divide-and-conquer algorithms for localization of adhoc sensor networks. The network is said to be uniquely registrable if the global coordinates can be computed uniquely (up to a rigid transform). Clearly, if the network is not uniquely registrable, then any registration algorithm whatsoever is bound to fail. We formulate a necessary and sufficient condition for uniquely registra-bility in arbitrary dimensions. This condition leads to a randomized polynomial-time test for unique registrability in arbitrary dimensions, and a combinatorial linear-time test in two dimensions.
{"title":"When Can a System of Subnetworks Be Registered Uniquely?","authors":"A. V. Singh, K. Chaudhury","doi":"10.1109/ICASSP.2019.8682680","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682680","url":null,"abstract":"Consider a network with N nodes in d dimensions, and M overlapping subsets P1, ⋯,PM (subnetworks). Assume that the nodes in a given Pi are observed in a local coordinate system. We wish to register the subnetworks using the knowledge of the observed coordinates. More precisely, we want to compute the positions of the N nodes in a global coordinate system, given P1, ⋯, PM and the corresponding local coordinates. Among other applications, this problem arises in divide-and-conquer algorithms for localization of adhoc sensor networks. The network is said to be uniquely registrable if the global coordinates can be computed uniquely (up to a rigid transform). Clearly, if the network is not uniquely registrable, then any registration algorithm whatsoever is bound to fail. We formulate a necessary and sufficient condition for uniquely registra-bility in arbitrary dimensions. This condition leads to a randomized polynomial-time test for unique registrability in arbitrary dimensions, and a combinatorial linear-time test in two dimensions.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"64 1","pages":"4564-4568"},"PeriodicalIF":0.0,"publicationDate":"2019-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84018914","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 : 2019-05-20DOI: 10.1109/ICASSP.2019.8682714
Onkar Krishna, Go Irie, Xiaomeng Wu, T. Kawanishi, K. Kashino
Finding a region of an image which matches to a query from a large number of candidates is a fundamental problem in image processing. The exhaustive nature of the sliding window approach has encouraged works that can reduce the run time by skipping unnecessary windows or pixels that do not play a substantial role in search results. However, such a pruning-based approach still needs to evaluate the non-ignorable number of candidates, which leads to a limited efficiency improvement. We propose an approach to learn efficient search paths from data. Our model is based on a CNN-LSTM architecture which is designed to sequentially determine a prospective location to be searched next based on the history of the locations attended. We propose a reinforcement learning algorithm to train the model in an end-to-end manner, which allows to jointly learn the search paths and deep image features for matching. These properties together significantly reduce the number of windows to be evaluated and makes it robust to background clutters. Our model gives remarkable matching accuracy with the reduced number of windows and run time on MNIST and FlickrLogos-32 datasets.
{"title":"Learning Search Path for Region-level Image Matching","authors":"Onkar Krishna, Go Irie, Xiaomeng Wu, T. Kawanishi, K. Kashino","doi":"10.1109/ICASSP.2019.8682714","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682714","url":null,"abstract":"Finding a region of an image which matches to a query from a large number of candidates is a fundamental problem in image processing. The exhaustive nature of the sliding window approach has encouraged works that can reduce the run time by skipping unnecessary windows or pixels that do not play a substantial role in search results. However, such a pruning-based approach still needs to evaluate the non-ignorable number of candidates, which leads to a limited efficiency improvement. We propose an approach to learn efficient search paths from data. Our model is based on a CNN-LSTM architecture which is designed to sequentially determine a prospective location to be searched next based on the history of the locations attended. We propose a reinforcement learning algorithm to train the model in an end-to-end manner, which allows to jointly learn the search paths and deep image features for matching. These properties together significantly reduce the number of windows to be evaluated and makes it robust to background clutters. Our model gives remarkable matching accuracy with the reduced number of windows and run time on MNIST and FlickrLogos-32 datasets.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"1 1","pages":"1967-1971"},"PeriodicalIF":0.0,"publicationDate":"2019-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88674951","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 : 2019-05-18DOI: 10.1109/ICASSP.2019.8682826
Selen Hande Kabil, Mathew Magimai Doss
Children speech recognition based on short-term spectral features is a challenging task. One of the reasons is that children speech has high fundamental frequency that is comparable to formant frequency values. Furthermore, as children grow, their vocal apparatus also undergoes changes. This presents difficulties in extracting standard short-term spectral-based features reliably for speech recognition. In recent years, novel acoustic modeling methods have emerged that learn both the feature and phone classifier in an end-to-end manner from the raw speech signal. Through an investigation on PF-STAR corpus we show that children speech recognition can be improved using end-to-end acoustic modeling methods.
{"title":"Improving Children Speech Recognition through Feature Learning from Raw Speech Signal","authors":"Selen Hande Kabil, Mathew Magimai Doss","doi":"10.1109/ICASSP.2019.8682826","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682826","url":null,"abstract":"Children speech recognition based on short-term spectral features is a challenging task. One of the reasons is that children speech has high fundamental frequency that is comparable to formant frequency values. Furthermore, as children grow, their vocal apparatus also undergoes changes. This presents difficulties in extracting standard short-term spectral-based features reliably for speech recognition. In recent years, novel acoustic modeling methods have emerged that learn both the feature and phone classifier in an end-to-end manner from the raw speech signal. Through an investigation on PF-STAR corpus we show that children speech recognition can be improved using end-to-end acoustic modeling methods.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"135 3 1","pages":"5736-5740"},"PeriodicalIF":0.0,"publicationDate":"2019-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82389424","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 : 2019-05-17DOI: 10.1109/ICASSP.2019.8682614
Takehiro Kono, M. Yukawa, Tomasz Piotrowski
We present a convexly-constrained beamformer design for brain activity reconstruction from non-invasive electroencephalography (EEG) signals. An intrinsic gap between the output variance and the mean squared errors is highlighted that occurs due to the presence of interfering activities correlated with the desired activity. The key idea of the proposed beamformer is reducing this gap without amplifying the noise by imposing a quadratic constraint that bounds the total power of interference leakage together with the distortionless constraint. The proposed beamformer can be implemented efficiently by the multi-domain adaptive filtering algorithm. Numerical examples show the clear advantages of the proposed beamformer over the minimum-variance distortionless response (MVDR) and nulling beamformers.
{"title":"Beamformer Design under Time-correlated Interference and Online Implementation: Brain-activity Reconstruction from EEG","authors":"Takehiro Kono, M. Yukawa, Tomasz Piotrowski","doi":"10.1109/ICASSP.2019.8682614","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8682614","url":null,"abstract":"We present a convexly-constrained beamformer design for brain activity reconstruction from non-invasive electroencephalography (EEG) signals. An intrinsic gap between the output variance and the mean squared errors is highlighted that occurs due to the presence of interfering activities correlated with the desired activity. The key idea of the proposed beamformer is reducing this gap without amplifying the noise by imposing a quadratic constraint that bounds the total power of interference leakage together with the distortionless constraint. The proposed beamformer can be implemented efficiently by the multi-domain adaptive filtering algorithm. Numerical examples show the clear advantages of the proposed beamformer over the minimum-variance distortionless response (MVDR) and nulling beamformers.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"35 1","pages":"1070-1074"},"PeriodicalIF":0.0,"publicationDate":"2019-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83722559","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 : 2019-05-17DOI: 10.1109/ICASSP.2019.8683669
Enea Ceolini, Jithendar Anumula, Stefan Braun, Shih-Chii Liu
This work presents an event-driven acoustic sensor processing pipeline to power a low-resource voice-activated smart assistant. The pipeline includes four major steps; namely localization, source separation, keyword spotting (KWS) and speaker verification (SV). The pipeline is driven by a front-end binaural spiking silicon cochlea sensor. The timing information carried by the output spikes of the cochlea provide spatial cues for localization and source separation. Spike features are generated with low latencies from the separated source spikes and are used by both KWS and SV which rely on state-of-the-art deep recurrent neural network architectures with a small memory footprint. Evaluation on a self-recorded event dataset based on TIDIGITS shows accuracies of over 93% and 88% on KWS and SV respectively, with minimum system latency of 5 ms on a limited resource device.
{"title":"Event-driven Pipeline for Low-latency Low-compute Keyword Spotting and Speaker Verification System","authors":"Enea Ceolini, Jithendar Anumula, Stefan Braun, Shih-Chii Liu","doi":"10.1109/ICASSP.2019.8683669","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8683669","url":null,"abstract":"This work presents an event-driven acoustic sensor processing pipeline to power a low-resource voice-activated smart assistant. The pipeline includes four major steps; namely localization, source separation, keyword spotting (KWS) and speaker verification (SV). The pipeline is driven by a front-end binaural spiking silicon cochlea sensor. The timing information carried by the output spikes of the cochlea provide spatial cues for localization and source separation. Spike features are generated with low latencies from the separated source spikes and are used by both KWS and SV which rely on state-of-the-art deep recurrent neural network architectures with a small memory footprint. Evaluation on a self-recorded event dataset based on TIDIGITS shows accuracies of over 93% and 88% on KWS and SV respectively, with minimum system latency of 5 ms on a limited resource device.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"1 1","pages":"7953-7957"},"PeriodicalIF":0.0,"publicationDate":"2019-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73201418","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 : 2019-05-16DOI: 10.1109/ICASSP.2019.8683366
Stephan Weiss, M. Macleod
This paper introduces a cost function for the smoothness of a continuous periodic function, of which only some samples are given. This cost function is important e.g. when associating samples in frequency bins for problems such as analytic singular or eigenvalue decompositions. We demonstrate the utility of the cost function, and study some of its complexity and conditioning issues.
{"title":"Maximally Smooth Dirichlet Interpolation from Complete and Incomplete Sample Points on the Unit Circle","authors":"Stephan Weiss, M. Macleod","doi":"10.1109/ICASSP.2019.8683366","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8683366","url":null,"abstract":"This paper introduces a cost function for the smoothness of a continuous periodic function, of which only some samples are given. This cost function is important e.g. when associating samples in frequency bins for problems such as analytic singular or eigenvalue decompositions. We demonstrate the utility of the cost function, and study some of its complexity and conditioning issues.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"84 1","pages":"8053-8057"},"PeriodicalIF":0.0,"publicationDate":"2019-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83857024","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 : 2019-05-16DOI: 10.1109/ICASSP.2019.8683500
B. M. Rafi, K. Murty
In this paper, the importance of analytic phase of the speech signal in automatic speaker verification systems is demonstrated in the context of replay spoof attacks. In order to accurately detect the replay spoof attacks, effective feature representations of speech signals are required to capture the distortion introduced due to the intermediate playback/recording devices, which is convolutive in nature. Since the convolutional distortion in time-domain translates to additive distortion in the phase-domain, we propose to use IFCC features extracted from the analytic phase of the speech signal. The IFCC features contain information from both clean speech and distortion components. The clean speech component has to be subtracted in order to highlight the distortion component introduced by the playback/recording devices. In this work, a dictionary learned from the IFCCs extracted from clean speech data is used to remove the clean speech component. The residual distortion component is used as a feature to build binary classifier for replay spoof detection. The proposed phase-based features delivered a 9% absolute improvement over the baseline system built using magnitude-based CQCC features.
{"title":"Importance of Analytic Phase of the Speech Signal for Detecting Replay Attacks in Automatic Speaker Verification Systems","authors":"B. M. Rafi, K. Murty","doi":"10.1109/ICASSP.2019.8683500","DOIUrl":"https://doi.org/10.1109/ICASSP.2019.8683500","url":null,"abstract":"In this paper, the importance of analytic phase of the speech signal in automatic speaker verification systems is demonstrated in the context of replay spoof attacks. In order to accurately detect the replay spoof attacks, effective feature representations of speech signals are required to capture the distortion introduced due to the intermediate playback/recording devices, which is convolutive in nature. Since the convolutional distortion in time-domain translates to additive distortion in the phase-domain, we propose to use IFCC features extracted from the analytic phase of the speech signal. The IFCC features contain information from both clean speech and distortion components. The clean speech component has to be subtracted in order to highlight the distortion component introduced by the playback/recording devices. In this work, a dictionary learned from the IFCCs extracted from clean speech data is used to remove the clean speech component. The residual distortion component is used as a feature to build binary classifier for replay spoof detection. The proposed phase-based features delivered a 9% absolute improvement over the baseline system built using magnitude-based CQCC features.","PeriodicalId":13203,"journal":{"name":"ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","volume":"16 1","pages":"6306-6310"},"PeriodicalIF":0.0,"publicationDate":"2019-05-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85227808","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：