Pub Date : 2013-11-01DOI: 10.1109/TASL.2013.2273715
Min Zhang, Wenliang Chen, Xiangyu Duan, Rong Zhang
For graph-based dependency parsing, how to enrich high-order features without increasing decoding complexity is a very challenging problem. To solve this problem, this paper presents an approach to representing high-order features for graph-based dependency parsing models using a dependency language model and beam search. Firstly, we use a baseline parser to parse a large-amount of unannotated data. Then we build the dependency language model (DLM) on the auto-parsed data. A set of new features is represented based on the DLM. Finally, we integrate the DLM-based features into the parsing model during decoding by beam search. We also utilize the features in bilingual text (bitext) parsing models. The main advantages of our approach are: 1) we utilize rich high-order features defined over a view of large scope and additional large raw corpus; 2) our approach does not increase the decoding complexity. We evaluate the proposed approach on the monotext and bitext parsing tasks. In the monotext parsing task, we conduct the experiments on Chinese and English data. The experimental results show that our new parser achieves the best accuracy on the Chinese data and comparable accuracy with the best known systems on the English data. In the bitext parsing task, we conduct the experiments on a Chinese-English bilingual data and our score is the best reported so far.
{"title":"Improving Graph-Based Dependency Parsing Models With Dependency Language Models","authors":"Min Zhang, Wenliang Chen, Xiangyu Duan, Rong Zhang","doi":"10.1109/TASL.2013.2273715","DOIUrl":"https://doi.org/10.1109/TASL.2013.2273715","url":null,"abstract":"For graph-based dependency parsing, how to enrich high-order features without increasing decoding complexity is a very challenging problem. To solve this problem, this paper presents an approach to representing high-order features for graph-based dependency parsing models using a dependency language model and beam search. Firstly, we use a baseline parser to parse a large-amount of unannotated data. Then we build the dependency language model (DLM) on the auto-parsed data. A set of new features is represented based on the DLM. Finally, we integrate the DLM-based features into the parsing model during decoding by beam search. We also utilize the features in bilingual text (bitext) parsing models. The main advantages of our approach are: 1) we utilize rich high-order features defined over a view of large scope and additional large raw corpus; 2) our approach does not increase the decoding complexity. We evaluate the proposed approach on the monotext and bitext parsing tasks. In the monotext parsing task, we conduct the experiments on Chinese and English data. The experimental results show that our new parser achieves the best accuracy on the Chinese data and comparable accuracy with the best known systems on the English data. In the bitext parsing task, we conduct the experiments on a Chinese-English bilingual data and our score is the best reported so far.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2273715","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891502","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 : 2013-11-01DOI: 10.1109/TASL.2013.2280209
Fabian Triefenbach, A. Jalalvand, Kris Demuynck, J. Martens
Accurate acoustic modeling is an essential requirement of a state-of-the-art continuous speech recognizer. The Acoustic Model (AM) describes the relation between the observed speech signal and the non-observable sequence of phonetic units uttered by the speaker. Nowadays, most recognizers use Hidden Markov Models (HMMs) in combination with Gaussian Mixture Models (GMMs) to model the acoustics, but neural-based architectures are on the rise again. In this work, the recently introduced Reservoir Computing (RC) paradigm is used for acoustic modeling. A reservoir is a fixed - and thus non-trained - Recurrent Neural Network (RNN) that is combined with a trained linear model. This approach combines the ability of an RNN to model the recent past of the input sequence with a simple and reliable training procedure. It is shown here that simple reservoir-based AMs achieve reasonable phone recognition and that deep hierarchical and bi-directional reservoir architectures lead to a very competitive Phone Error Rate (PER) of 23.1% on the well-known TIMIT task.
{"title":"Acoustic Modeling With Hierarchical Reservoirs","authors":"Fabian Triefenbach, A. Jalalvand, Kris Demuynck, J. Martens","doi":"10.1109/TASL.2013.2280209","DOIUrl":"https://doi.org/10.1109/TASL.2013.2280209","url":null,"abstract":"Accurate acoustic modeling is an essential requirement of a state-of-the-art continuous speech recognizer. The Acoustic Model (AM) describes the relation between the observed speech signal and the non-observable sequence of phonetic units uttered by the speaker. Nowadays, most recognizers use Hidden Markov Models (HMMs) in combination with Gaussian Mixture Models (GMMs) to model the acoustics, but neural-based architectures are on the rise again. In this work, the recently introduced Reservoir Computing (RC) paradigm is used for acoustic modeling. A reservoir is a fixed - and thus non-trained - Recurrent Neural Network (RNN) that is combined with a trained linear model. This approach combines the ability of an RNN to model the recent past of the input sequence with a simple and reliable training procedure. It is shown here that simple reservoir-based AMs achieve reasonable phone recognition and that deep hierarchical and bi-directional reservoir architectures lead to a very competitive Phone Error Rate (PER) of 23.1% on the well-known TIMIT task.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2280209","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62892491","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 : 2013-11-01DOI: 10.1109/TASL.2013.2273716
Florian Pflug, T. Fingscheidt
Applications such as professional wireless digital microphones require a transmission of practically uncoded high-quality audio with ultra-low latency on the one hand and robustness to error-prone channels on the other hand. The delay restrictions, however, prohibit the utilization of efficient block or convolutional channel codes for error protection. The contribution of this work is fourfold: We revise and summarize concisely a Bayesian framework for soft-decision audio decoding and present three novel approaches to (almost) latency-free robust decoding of uncompressed audio. Bit reliability information from the transmission channel is exploited, as well as short-term and long-term residual redundancy within the audio signal, and optionally some explicit redundancy in terms of a sample-individual block code. In all cases we utilize variants of higher-order linear prediction to compute prediction probabilities in three novel ways: Firstly by employing a serial cascade of multiple predictors, secondly by exploiting explicit redundancy in form of parity bits, and thirdly by utilizing an interpolative forward/backward prediction algorithm. The first two presented approaches work fully delayless, while the third one introduces an ultra-low algorithmic delay of just a few samples. The effectiveness of the proposed algorithms is proven in simulations with BPSK and typical digital microphone FSK modulation schemes on AWGN and bursty fading channels.
{"title":"Robust Ultra-Low Latency Soft-Decision Decoding of Linear PCM Audio","authors":"Florian Pflug, T. Fingscheidt","doi":"10.1109/TASL.2013.2273716","DOIUrl":"https://doi.org/10.1109/TASL.2013.2273716","url":null,"abstract":"Applications such as professional wireless digital microphones require a transmission of practically uncoded high-quality audio with ultra-low latency on the one hand and robustness to error-prone channels on the other hand. The delay restrictions, however, prohibit the utilization of efficient block or convolutional channel codes for error protection. The contribution of this work is fourfold: We revise and summarize concisely a Bayesian framework for soft-decision audio decoding and present three novel approaches to (almost) latency-free robust decoding of uncompressed audio. Bit reliability information from the transmission channel is exploited, as well as short-term and long-term residual redundancy within the audio signal, and optionally some explicit redundancy in terms of a sample-individual block code. In all cases we utilize variants of higher-order linear prediction to compute prediction probabilities in three novel ways: Firstly by employing a serial cascade of multiple predictors, secondly by exploiting explicit redundancy in form of parity bits, and thirdly by utilizing an interpolative forward/backward prediction algorithm. The first two presented approaches work fully delayless, while the third one introduces an ultra-low algorithmic delay of just a few samples. The effectiveness of the proposed algorithms is proven in simulations with BPSK and typical digital microphone FSK modulation schemes on AWGN and bursty fading channels.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2273716","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891352","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 : 2013-11-01DOI: 10.1109/TASL.2013.2274694
Koji Seto, T. Ogunfunmi
High quality speech at low bit rates makes code excited linear prediction (CELP) the dominant choice for a narrowband coding technique despite the susceptibility to packet loss. One of the few techniques which received attention after the introduction of CELP coding technique is the internet low bitrate codec (iLBC) because of inherent high robustness to packet loss. Addition of rate flexibility and scalability makes the iLBC an attractive choice for voice communication over IP networks. In this paper, performance improvement schemes of multi-rate iLBC and its scalable structure are proposed, and the proposed codec enhanced from the previous work is re-designed based on the subjective listening quality instead of the objective quality. In particular, perceptual weighting and the modified discrete cosine transform (MDCT) with short overlap in weighted signal domain are employed along with the improved packet loss concealment (PLC) algorithm. The subjective evaluation results show that the speech quality of the proposed codec is equivalent to that of state-of-the-art codec, G.718, under both a clean channel condition and lossy channel conditions. This result is significant considering that development of the proposed codec is still in early stage.
{"title":"Scalable Speech Coding for IP Networks: Beyond iLBC","authors":"Koji Seto, T. Ogunfunmi","doi":"10.1109/TASL.2013.2274694","DOIUrl":"https://doi.org/10.1109/TASL.2013.2274694","url":null,"abstract":"High quality speech at low bit rates makes code excited linear prediction (CELP) the dominant choice for a narrowband coding technique despite the susceptibility to packet loss. One of the few techniques which received attention after the introduction of CELP coding technique is the internet low bitrate codec (iLBC) because of inherent high robustness to packet loss. Addition of rate flexibility and scalability makes the iLBC an attractive choice for voice communication over IP networks. In this paper, performance improvement schemes of multi-rate iLBC and its scalable structure are proposed, and the proposed codec enhanced from the previous work is re-designed based on the subjective listening quality instead of the objective quality. In particular, perceptual weighting and the modified discrete cosine transform (MDCT) with short overlap in weighted signal domain are employed along with the improved packet loss concealment (PLC) algorithm. The subjective evaluation results show that the speech quality of the proposed codec is equivalent to that of state-of-the-art codec, G.718, under both a clean channel condition and lossy channel conditions. This result is significant considering that development of the proposed codec is still in early stage.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2274694","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891744","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 : 2013-11-01DOI: 10.1109/TASL.2013.2277928
Symeon Delikaris-Manias, V. Pulkki
A parametric spatial filtering algorithm with a fixed beam direction is proposed in this paper. The algorithm utilizes the normalized cross-spectral density between signals from microphones of different orders as a criterion for focusing in specific directions. The correlation between microphone signals is estimated in the time-frequency domain. A post-filter is calculated from a multichannel input and is used to assign attenuation values to a coincidentally captured audio signal. The proposed algorithm is simple to implement and offers the capability of coping with interfering sources at different azimuthal locations with or without the presence of diffuse sound. It is implemented by using directional microphones placed in the same look direction and have the same magnitude and phase response. Experiments are conducted with simulated and real microphone arrays employing the proposed post-filter and compared to previous coherence-based approaches, such as the McCowan post-filter. A significant improvement is demonstrated in terms of objective quality measures. Formal listening tests conducted to assess the audibility of artifacts of the proposed algorithm in real acoustical scenarios show that no annoying artifacts existed with certain spectral floor values. Examples of the proposed algorithm can be found online at http://www.acoustics.hut.fi/projects/cropac/soundExamples.
{"title":"Cross Pattern Coherence Algorithm for Spatial Filtering Applications Utilizing Microphone Arrays","authors":"Symeon Delikaris-Manias, V. Pulkki","doi":"10.1109/TASL.2013.2277928","DOIUrl":"https://doi.org/10.1109/TASL.2013.2277928","url":null,"abstract":"A parametric spatial filtering algorithm with a fixed beam direction is proposed in this paper. The algorithm utilizes the normalized cross-spectral density between signals from microphones of different orders as a criterion for focusing in specific directions. The correlation between microphone signals is estimated in the time-frequency domain. A post-filter is calculated from a multichannel input and is used to assign attenuation values to a coincidentally captured audio signal. The proposed algorithm is simple to implement and offers the capability of coping with interfering sources at different azimuthal locations with or without the presence of diffuse sound. It is implemented by using directional microphones placed in the same look direction and have the same magnitude and phase response. Experiments are conducted with simulated and real microphone arrays employing the proposed post-filter and compared to previous coherence-based approaches, such as the McCowan post-filter. A significant improvement is demonstrated in terms of objective quality measures. Formal listening tests conducted to assess the audibility of artifacts of the proposed algorithm in real acoustical scenarios show that no annoying artifacts existed with certain spectral floor values. Examples of the proposed algorithm can be found online at http://www.acoustics.hut.fi/projects/cropac/soundExamples.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2277928","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891892","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 : 2013-11-01DOI: 10.1109/TASLP.2013.2286921
Pasi Pertilä, M. Hämäläinen, Mikael Mieskolainen
In recent years ad-hoc microphone arrays have become ubiquitous, and the capture hardware and quality is increasingly more sophisticated. Ad-hoc arrays hold a vast potential for audio applications, but they are inherently asynchronous, i.e., temporal offset exists in each channel, and furthermore the device locations are generally unknown. Therefore, the data is not directly suitable for traditional microphone array applications such as source localization and beamforming. This work presents a least squares method for temporal offset estimation of a static ad-hoc microphone array. The method utilizes the captured audio content without the need to emit calibration signals, provided that during the recording a sufficient amount of sound sources surround the array. The Cramer-Rao lower bound of the estimator is given and the effect of limited number of surrounding sources on the solution accuracy is investigated. A practical implementation is then presented using non-linear filtering with automatic parameter adjustment. Simulations over a range of reverberation and noise levels demonstrate the algorithm's robustness. Using smartphones an average RMS error of 3.5 samples (at 48 kHz) was reached when the algorithm's assumptions were met.
{"title":"Passive Temporal Offset Estimation of Multichannel Recordings of an Ad-Hoc Microphone Array","authors":"Pasi Pertilä, M. Hämäläinen, Mikael Mieskolainen","doi":"10.1109/TASLP.2013.2286921","DOIUrl":"https://doi.org/10.1109/TASLP.2013.2286921","url":null,"abstract":"In recent years ad-hoc microphone arrays have become ubiquitous, and the capture hardware and quality is increasingly more sophisticated. Ad-hoc arrays hold a vast potential for audio applications, but they are inherently asynchronous, i.e., temporal offset exists in each channel, and furthermore the device locations are generally unknown. Therefore, the data is not directly suitable for traditional microphone array applications such as source localization and beamforming. This work presents a least squares method for temporal offset estimation of a static ad-hoc microphone array. The method utilizes the captured audio content without the need to emit calibration signals, provided that during the recording a sufficient amount of sound sources surround the array. The Cramer-Rao lower bound of the estimator is given and the effect of limited number of surrounding sources on the solution accuracy is investigated. A practical implementation is then presented using non-linear filtering with automatic parameter adjustment. Simulations over a range of reverberation and noise levels demonstrate the algorithm's robustness. Using smartphones an average RMS error of 3.5 samples (at 48 kHz) was reached when the algorithm's assumptions were met.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASLP.2013.2286921","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62892231","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 : 2013-11-01DOI: 10.1109/TASL.2013.2263142
Gillian M. Chin, J. Nocedal, P. Olsen, Steven J. Rennie
A variety of first-order methods have recently been proposed for solving matrix optimization problems arising in machine learning. The premise for utilizing such algorithms is that second order information is too expensive to employ, and so simple first-order iterations are likely to be optimal. In this paper, we argue that second-order information is in fact efficiently accessible in many matrix optimization problems, and can be effectively incorporated into optimization algorithms. We begin by reviewing how certain Hessian operations can be conveniently represented in a wide class of matrix optimization problems, and provide the first proofs for these results. Next we consider a concrete problem, namely the minimization of the ℓ1 regularized Jeffreys divergence, and derive formulae for computing Hessians and Hessian vector products. This allows us to propose various second order methods for solving the Jeffreys divergence problem. We present extensive numerical results illustrating the behavior of the algorithms and apply the methods to a speech recognition problem. We compress full covariance Gaussian mixture models utilized for acoustic models in automatic speech recognition. By discovering clusters of (sparse inverse) covariance matrices, we can compress the number of covariance parameters by a factor exceeding 200, while still outperforming the word error rate (WER) performance of a diagonal covariance model that has 20 times less covariance parameters than the original acoustic model.
{"title":"Second Order Methods for Optimizing Convex Matrix Functions and Sparse Covariance Clustering","authors":"Gillian M. Chin, J. Nocedal, P. Olsen, Steven J. Rennie","doi":"10.1109/TASL.2013.2263142","DOIUrl":"https://doi.org/10.1109/TASL.2013.2263142","url":null,"abstract":"A variety of first-order methods have recently been proposed for solving matrix optimization problems arising in machine learning. The premise for utilizing such algorithms is that second order information is too expensive to employ, and so simple first-order iterations are likely to be optimal. In this paper, we argue that second-order information is in fact efficiently accessible in many matrix optimization problems, and can be effectively incorporated into optimization algorithms. We begin by reviewing how certain Hessian operations can be conveniently represented in a wide class of matrix optimization problems, and provide the first proofs for these results. Next we consider a concrete problem, namely the minimization of the ℓ1 regularized Jeffreys divergence, and derive formulae for computing Hessians and Hessian vector products. This allows us to propose various second order methods for solving the Jeffreys divergence problem. We present extensive numerical results illustrating the behavior of the algorithms and apply the methods to a speech recognition problem. We compress full covariance Gaussian mixture models utilized for acoustic models in automatic speech recognition. By discovering clusters of (sparse inverse) covariance matrices, we can compress the number of covariance parameters by a factor exceeding 200, while still outperforming the word error rate (WER) performance of a diagonal covariance model that has 20 times less covariance parameters than the original acoustic model.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2263142","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62889848","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 : 2013-11-01DOI: 10.1109/TASL.2013.2271592
Theodoros Tsiligkaridis, E. Marcheret, V. Goel
We introduce a new class of parameter estimation methods for log-linear models. Our approach relies on the fact that minimizing a rational function of mixtures of exponentials is equivalent to minimizing a difference of convex functions. This allows us to construct convex auxiliary functions by applying the concave-convex procedure (CCCP). We consider a modification of CCCP where a proximal term is added (ProxCCCP), and extend it further by introducing an ℓ1 penalty. For solving the ` convex + ℓ1' auxiliary problem, we propose an approach called SeqGPSR that is based on sequential application of the GPSR procedure. We present convergence analysis of the algorithms, including sufficient conditions for convergence to a critical point of the objective function. We propose an adaptive procedure for varying the strength of the proximal regularization term in each ProxCCCP iteration, and show this procedure (AProxCCCP) is effective in practice and stable under some mild conditions. The CCCP procedure and proposed variants are applied to the task of optimizing the cross-entropy objective function for an audio frame classification problem. Class posteriors are modeled using log-linear models consisting of approximately 6 million parameters. Our results show that CCCP variants achieve a much better cross-entropy objective value as compared to direct optimization of the objective function by a first order gradient based approach, stochastic gradient descent or the L-BFGS procedure.
{"title":"A Difference of Convex Functions Approach to Large-Scale Log-Linear Model Estimation","authors":"Theodoros Tsiligkaridis, E. Marcheret, V. Goel","doi":"10.1109/TASL.2013.2271592","DOIUrl":"https://doi.org/10.1109/TASL.2013.2271592","url":null,"abstract":"We introduce a new class of parameter estimation methods for log-linear models. Our approach relies on the fact that minimizing a rational function of mixtures of exponentials is equivalent to minimizing a difference of convex functions. This allows us to construct convex auxiliary functions by applying the concave-convex procedure (CCCP). We consider a modification of CCCP where a proximal term is added (ProxCCCP), and extend it further by introducing an ℓ1 penalty. For solving the ` convex + ℓ1' auxiliary problem, we propose an approach called SeqGPSR that is based on sequential application of the GPSR procedure. We present convergence analysis of the algorithms, including sufficient conditions for convergence to a critical point of the objective function. We propose an adaptive procedure for varying the strength of the proximal regularization term in each ProxCCCP iteration, and show this procedure (AProxCCCP) is effective in practice and stable under some mild conditions. The CCCP procedure and proposed variants are applied to the task of optimizing the cross-entropy objective function for an audio frame classification problem. Class posteriors are modeled using log-linear models consisting of approximately 6 million parameters. Our results show that CCCP variants achieve a much better cross-entropy objective value as compared to direct optimization of the objective function by a first order gradient based approach, stochastic gradient descent or the L-BFGS procedure.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2271592","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891324","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 : 2013-11-01DOI: 10.1109/TASL.2013.2271591
P. Cardinal, P. Dumouchel, Gilles Boulianne
The speed of modern processors has remained constant over the last few years but the integration capacity continues to follow Moore's law and thus, to be scalable, applications must be parallelized. The parallelization of the classical Viterbi beam search has been shown to be very difficult on multi-core processor architectures or massively threaded architectures such as Graphics Processing Unit (GPU). The problem with this approach is that active states are scattered in memory and thus, they cannot be efficiently transferred to the processor memory. This problem can be circumvented by using the A* search which uses a heuristic to significantly reduce the number of explored hypotheses. The main advantage of this algorithm is that the processing time is moved from the search in the recognition network to the computation of heuristic costs, which can be designed to take advantage of parallel architectures. Our parallel implementation of the A* decoder on a 4-core processor with a GPU led to a speed-up factor of 6.13 compared to the Viterbi beam search at its maximum capacity and an improvement of 4% absolute in accuracy at real-time.
{"title":"Large Vocabulary Speech Recognition on Parallel Architectures","authors":"P. Cardinal, P. Dumouchel, Gilles Boulianne","doi":"10.1109/TASL.2013.2271591","DOIUrl":"https://doi.org/10.1109/TASL.2013.2271591","url":null,"abstract":"The speed of modern processors has remained constant over the last few years but the integration capacity continues to follow Moore's law and thus, to be scalable, applications must be parallelized. The parallelization of the classical Viterbi beam search has been shown to be very difficult on multi-core processor architectures or massively threaded architectures such as Graphics Processing Unit (GPU). The problem with this approach is that active states are scattered in memory and thus, they cannot be efficiently transferred to the processor memory. This problem can be circumvented by using the A* search which uses a heuristic to significantly reduce the number of explored hypotheses. The main advantage of this algorithm is that the processing time is moved from the search in the recognition network to the computation of heuristic costs, which can be designed to take advantage of parallel architectures. Our parallel implementation of the A* decoder on a 4-core processor with a GPU led to a speed-up factor of 6.13 compared to the Viterbi beam search at its maximum capacity and an improvement of 4% absolute in accuracy at real-time.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2271591","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891565","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 : 2013-11-01DOI: 10.1109/TASL.2013.2274695
K. Niwa, Yusuke Hioka, K. Furuya, Y. Haneda
We generalized our previously proposed diffused sensing for a microphone array design to achieve sharp directive beamforming to enable various filter design methods to be applied. In the conventional microphone array, various filter design methods have been studied to narrow the directivity beam width. However, it is difficult to minimize the power of interference sources in the beamforming output (output interference power) over a broad frequency range since the cross-correlation between transfer functions from sound sources to microphones increases in some frequencies. With the diffused sensing, the cross-correlation is minimized by physically varying the transfer functions. We investigated how a microphone array should be designed in order to minimize the cross-correlation between transfer functions and found that placing the array in a diffuse acoustic field produces optimum results. Because the transfer functions are known a priori, this finding makes it possible to narrow the directivity beam width over a broad frequency range. This observation can be practically achieved by placing microphones inside a reflective enclosure, part of which is open to let sound waves enter. We conducted experiments using 24 microphones and confirmed that the output interference power was reduced over a broad frequency range and the beam width was narrowed by using the diffused sensing.
{"title":"Diffused Sensing for Sharp Directive Beamforming","authors":"K. Niwa, Yusuke Hioka, K. Furuya, Y. Haneda","doi":"10.1109/TASL.2013.2274695","DOIUrl":"https://doi.org/10.1109/TASL.2013.2274695","url":null,"abstract":"We generalized our previously proposed diffused sensing for a microphone array design to achieve sharp directive beamforming to enable various filter design methods to be applied. In the conventional microphone array, various filter design methods have been studied to narrow the directivity beam width. However, it is difficult to minimize the power of interference sources in the beamforming output (output interference power) over a broad frequency range since the cross-correlation between transfer functions from sound sources to microphones increases in some frequencies. With the diffused sensing, the cross-correlation is minimized by physically varying the transfer functions. We investigated how a microphone array should be designed in order to minimize the cross-correlation between transfer functions and found that placing the array in a diffuse acoustic field produces optimum results. Because the transfer functions are known a priori, this finding makes it possible to narrow the directivity beam width over a broad frequency range. This observation can be practically achieved by placing microphones inside a reflective enclosure, part of which is open to let sound waves enter. We conducted experiments using 24 microphones and confirmed that the output interference power was reduced over a broad frequency range and the beam width was narrowed by using the diffused sensing.","PeriodicalId":55014,"journal":{"name":"IEEE Transactions on Audio Speech and Language Processing","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2013-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1109/TASL.2013.2274695","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"62891867","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}