检索结果-内蒙古大学图书馆

IEEE/SP Workshop on Statistical Signal processing (SSP)

作者： Shuai Huang Damianos Karakos Daguang Xu Center of Language and Speech Processing Department of Electrical and Computer Engineering Johns Hopkins University Baltimore MD USA Human Language Technology Center of Excellence Center of Language and Speech Processing Department of Electrical and Computer Engineering Johns Hopkins University Baltimore MD USA

In recent work, Lyu and Simoncelli [1] introduced radial Gaussianization (RG) as a very efficient procedure for transforming n-dimensional random vectors into Gaussian vectors with independent and identically distributed (i.i.d.) components. This entails transforming the norms of the data so that they become chi-distributed with n degrees of freedom. A necessary requirement is that the original data are generated by an isotropic distribution, that is, their probability density function (pdf) is constant over surfaces of n-dimensional spheres (or, more general, n-dimensional ellipsoids). The case of biases in the data, which is of great practical interest, is studied here; as we demonstrate with experiments, there are situations in which even very small amounts of bias can cause RG to fail. This becomes evident especially when the data form clusters in low-dimensional manifolds. To address this shortcoming, we propose a two-step approach which entails (i) first discovering clusters in the data and removing the bias from each, and (ii) performing RG on the bias-compensated data. In experiments with synthetic data, the proposed bias compensation procedure results in significantly better Gaus-sianization than the non-compensated RG method.

关键词： Principal component analysis Noise measurement Estimation Gaussian distribution Distributed databases Optimization Transforms

来源：评论

学校读者我要写书评

暂无评论

MULTILAYER PERCEPTRON WITH SPARSE HIDDEN OUTPUTS FOR PHONEME RECOGNITION

MULTILAYER PERCEPTRON WITH SPARSE HIDDEN OUTPUTS FOR PHONEME...

引用

IEEE International Conference on Acoustics, speech and Signal processing

作者： G.S.V.S. Sivaram Hynek Hermansky Department of Electrical & Computer Engineering Center of Language and Speech Processing Human Language Technology Center of Excellence Johns Hopkins University USA

This paper introduces the sparse multilayer perceptron (SMLP) which learns the transformation from the inputs to the targets as in multilayer perceptron (MLP) while the outputs of one of the internal hidden layers is forced to be sparse. This is achieved by adding a sparse regularization term to the cross-entropy cost and learning the parameters of the network to minimize the joint cost. On the TIMIT phoneme recognition task, the SMLP based system trained using perceptual linear prediction (PLP) features performs better than the conventional MLP based system. Furthermore, their combination yields a phoneme error rate of 21.2percent, a relative improvement of 6.2percent over the baseline.

关键词： Hidden Markov models Neurons Training Cost function speech speech processing Acoustics

来源：评论

学校读者我要写书评

暂无评论

Multistream robust speaker recognition based on speech intelligibility

Multistream robust speaker recognition based on speech intel...

引用

Annual Conference on Information Sciences and Systems (CISS)

作者： Sridhar Krishna Nemala Mounya Elhilali Department of Electrical and Computer Engineering Center for Speech and Language Processing Johns Hopkins University Baltimore MD USA

Delimiting the most informative voice segments of an acoustic signal is often a crucial initial step for any speech processing system. In the current work, we propose a novel segmentation approach based on a perception-based measure of speech intelligibility. Unlike segmentation approaches based on various forms of voice-activity detection (VAD), the proposed segmentation approach exploits higher-level perceptual information about the signal intelligibility levels. This classification based on intelligibility estimates is integrated into a novel multistream framework for automatic speaker recognition task. The multistream system processes the input acoustic signal along multiple independent streams reflecting various levels of intelligibility and then fusing the decision scores from the multiple steams according to their intelligibility contribution. Our results show that the proposed multistream system achieves significant improvements both in clean and noisy conditions when compared with a baseline and a state-of-the-art voice-activity detection algorithm.

关键词： Computational modeling Robustness Noise

来源：评论

学校读者我要写书评

暂无评论

A comparative study of different classifier selection and fusion algorithms in a multimodal identity verification system

A comparative study of different classifier selection and fu...

引用

作者： Sedighi, Vahid Sadeghi, Mohammad T. Kittler, Josef Signal Processing Research Group Department of Electrical and Computer Engineering Yazd University Yazd Iran Center for Vision Speech and Signal Processing University of Surrey Guildford Surrey United Kingdom

ISBN: (纸本)9789898425478

A comparative study of a set of classifier selection and fusion algorithms for combining face and speech modalities in a multimodal biometric system is performed. Our text independent speaker verification system uses MFCC features along with a GMM method for speaker modelling and the GMM-UBM classifier for matching. The face verification system uses different similarity measures in two appearance based representation spaces namely PCA and LDA. The adopted authentication systems are fused at the matching score level to realise a system which can meet more challenging and varying requirements. The utility of two classifier selection methods, namely the sequential search method and Particle Swarm Optimisation (PSO) algorithm, along with two matching score fusion rules, the weighted averaging rule and SVM, is experimentally studied. In our extensive experimentation on the BANCA database, we demonstrate that by combining the PSO classifier selection method and the SVM fusion rule, the performance of the resulting decision making system is consistently superior.

关键词： Particle swarm optimization (PSO)

来源：评论

学校读者我要写书评

暂无评论

DIRICHLET MIXTURE MODELS OF NEURAL NET POSTERIORS FOR HMM-BASED speech RECOGNITION

DIRICHLET MIXTURE MODELS OF NEURAL NET POSTERIORS FOR HMM-BA...

引用

IEEE International Conference on Acoustics, speech and Signal processing

作者： Balakrishnan V G.S.V.S. Sivaram Sanjeev Khudanpur Dept. of Electrical & Computer Engineering Center for Language and Speech Processing The Johns Hopkins University USA

ISBN: (纸本)9781457705380

In this paper, we present a novel technique for modeling the posterior probability estimates obtained from a neural network directly in the HMM framework using the Dirichlet Mixture Models (DMMs). Since posterior probability vectors lie on a probability simplex their distribution can be modeled using DMMs. Being in an exponential family, the parameters of DMMs can be estimated in an efficient manner. Conventional approaches like TANDEM attempt to gaussianize the posteriors by suitable transforms and model them using Gaussian Mixture Models (GMMs). This requires more number of parameters as it does not exploit the fact that the probability vectors lie on a simplex. We demonstrate through TIMIT phoneme recognition experiments that the proposed technique outperforms the conventional TANDEM approach.

关键词： Dirichlet distribution neural network posteriors HMMs

来源：评论

学校读者我要写书评

暂无评论

Exploiting temporal coherence in speech for data-driven feature extraction

Exploiting temporal coherence in speech for data-driven feat...

引用

Annual Conference on Information Sciences and Systems

作者： Carlin, Michael A. Elhilali, Mounya Center for Language and Speech Processing Johns Hopkins University Baltimore MD 21218 United States Human Language Technology Center of Excellence Johns Hopkins University Baltimore MD 21218 United States Dept. of Electrical and Computer Engineering Johns Hopkins University Baltimore MD 21218 United States

ISBN: (纸本)9781424498475

It is well known that speech sounds evolve at multiple timescales over the course of tens to hundreds of milliseconds. Such temporal modulations are crucial for speech perception and are believed to directly influence the underlying code for representing acoustic stimuli. The present work seeks to explicitly quantify this relationship using the principle of temporal coherence. Here we show that by constraining the outputs of model linear neurons to be highly correlated over timescales relevant to speech, we observe the emergence of neural response fields that are bandpass, localized, and reflective of the rich spectro-temporal structure present in speech. The emergent response fields also appear to share qualitative similarities those observed in auditory neurophysiology. Importantly, learning is accomplished using unlabeled speech data, and the emergent neural properties well-characterize the spectro-temporal statistics of the input. We analyze the characteristics and coverage of ensembles of learned response fields for a variety of timescales, and suggest uses of such a coherence learning framework for common speech tasks. © 2011 IEEE.

关键词： speech

来源：评论

学校读者我要写书评

暂无评论

Randomized maximum entropy language models

Randomized maximum entropy language models

引用

IEEE Workshop on Automatic speech Recognition and Understanding

作者： Puyang Xu Sanjeev Khudanpur Asela Gunawardana Department of Electrical & Computer Engineering Center of Language and Speech Processing Johns Hopkins University Baltimore MD USA Microsoft Research Redmond WA USA

We address the memory problem of maximum entropy language models (MELM) with very large feature sets. Randomized techniques are employed to remove all large, exact data structures in MELM implementations. To avoid the dictionary structure that maps each feature to its corresponding weight, the feature hashing trick [1] [2] can be used. We also replace the explicit storage of features with a Bloom filter. We show with extensive experiments that false positive errors of Bloom filters and random hash collisions do not degrade model performance. Both perplexity and WER improvements are demonstrated by building MELM that would otherwise be prohibitively large to estimate or store.

关键词： Vectors Training Computational modeling Entropy Dictionaries Memory management Vocabulary

来源：评论

学校读者我要写书评

暂无评论

Adaptive variable degree-κ zero-trees for re-encoding of perceptually quantized wavelet packet transformed audio and high-quality speech

引用

ISRN Signal processing 2011年第1期2011卷

作者： Ghahabi, Omid Savoji, Mohammad Hassan Department of Electrical and Computer Engineering Shahid Beheshti University Evin Sq. Tehran *** Iran Audio and Speech Processing Group Research Center of Intelligent Signal Processing (RCISP) Tehran *** Iran

A fast, efficient, and scalable algorithm is proposed, in this paper, for re-encoding of perceptually quantized wavelet-packet transform (WPT) coefficients of audio and high quality speech and is called "adaptive variable degree-κ zero-trees" (AVDZ). The quantization process is carried out by taking into account some basic perceptual considerations and achieves good subjective quality with low complexity. The performance of the proposed AVDZ algorithm is compared with two other zero-tree-based schemes comprising (1) embedded zero-tree wavelet (EZW) and (2) the set partitioning in hierarchical trees (SPIHT). Since EZW and SPIHT are designed for image compression, some modifications are incorporated in these schemes for their better matching to audio signals. It is shown that the proposed modifications can improve their performance by about 15-25%. Furthermore, it is concluded that the proposed AVDZ algorithm outperforms these modified versions in terms of both output average bit-rates and computation times. Copyright © 2011 O. Ghahabi and M. H. Savoji.

关键词：

来源：评论

学校读者我要写书评

暂无评论

HYPOTHESIS RANKING AND TWO-PASS APPROACHES FOR MACHINE TRANSLATION SYSTEM COMBINATION

HYPOTHESIS RANKING AND TWO-PASS APPROACHES FOR MACHINE TRANS...

引用

IEEE International Conference on Acoustics, speech, and Signal processing

作者： Damianos Karakos Jason Smith Sanjeev Khudanpur Center for Language and Speech Processing Department of Electrical and Computer Engineering Johns Hopkins University Baltimore MD 21218 Center for Language and Speech Processing Department of Computer Science Johns Hopkins University Baltimore MD 21218

ISBN: (纸本)9781424442959

Given a number of machine translations of a source segment, the goal of system combination is to produce a new translation that has better quality than all of them. This paper describes a number of improvements that were recently added to the JHU system combination scheme: (i) A hypothesis ranking technique which orders the system outputs, on a per-segment basis, according to predicted translation quality, thus improving a subsequent incremental combination step. (ii) A two-pass combination procedure, which first produces several combination outputs with the given translations, and then performs one more combination step with these new outputs. Results from the NIST MT09 informal system combination evaluation on Arabic-to-English and Urdu-to-English show that both approaches offer significant BLEU and TER gains over a baseline JHU combination scheme.

关键词： Machine Translation System Combination Confusion Network Decoding Hypothesis Ranking

来源：评论

学校读者我要写书评

暂无评论

Ecological loudness: Binarual loudness constancy International Congress on Acoustics, ICA 2010

Ecological loudness: Binarual loudness constancy Internation...

引用

20th International Congress on Acoustics 2010, ICA 2010 - Incorporating the 2010 Annual Conference of the Australian Acoustical Society

作者： Florentine, Mary Epstein, Michael Dept. of Speech-Language Pathology and Audiology Institute for Hearing Speech and Language Northeastern University 360 Huntington Ave. Boston MA 02115 United States Auditory Modeling and Processing Laboratory Dept. of Speech-Language Pathology and Audiology Northeastern University 360 Huntington Ave Boston MA 02115 United States CDSP Center Dept. of Electrical and Computer Engineering Northeastern University 360 Huntington Ave Boston MA 02115 United States

ISBN: (纸本)9781617827457

Are conclusions about loudness drawn from tones presented via earphones in laboratories applicable to listening to a talker in a room? The present experiment tests the following hypothesis: speech from the same talkers presented under more ecologically valid conditions results in a smaller binaural-to-monaural loudness ratio than speech presented without visual cues and/or presented via headphones. Twelve normal listeners were presented two types of stimuli (recorded speech, with and without visual cues) monaurally and binaurally across a wide range of levels. The same stimuli were presented via earphones and loudspeakers. Loudness was measured using magnitude estimation. Results show that the binaural-to-monaural loudness ratio was significantly less for speech with visual cues presented via a loudspeaker than for stimuli with any other combination of test parameters (i.e., speech without visual cues presentedvia both headphones and loudspeakers, and speech presented with visual cues via headphones). The present results indicate that the loudness of a visually present talker in daily environments is little affected by switching between binaural and monaural listening. This phenomenon has been dubbed "Binaural Loudness Constancy," because of its similarity to loudness constancy that occurs with distance from the speaker. The present experiment supports the importance of ecological validity in loudness research, which could change how perception of loudness is understood.

关键词： speech

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：