检索结果-内蒙古大学图书馆

Intra-topic variability normalization based on linear projection for topic classification 15

Intra-topic variability normalization based on linear projec...

15th Conference of the North American Chapter of the Association for Computational Linguistics: Human language Technologies, NAACL HLT 2016

作者： Liu, Quan Guo, Wu Ling, Zhen-Hua Jiang, Hui Hu, Yu National Engineering Laboratory for Speech and Language Information Processing University of Science and Technology of China HefeiAnhui China Department of Electrical Engineering and Computer Science York University 4700 Keele Street TorontoONM3J 1P3 Canada iFLYTEK Research Hefei China

ISBN: (纸本)9781941643914

This paper proposes a variability normalization algorithm to reduce the variability between intra-topic documents for topic classification. Firstly, an optimization problem is constructed based on linear variability removable assumption. Secondly, a new feature space for document representation is found by solving the optimization problem with kernel principle component analysis (KPCA). Finally, effective feature transformation is taken through linear projection. As for experiments, state-of-the-art SVM and KNN algorithm are adopted for topic classification respectively. Experimental results on a free-style conversational corpus show that the proposed variability normalization algorithm for topic classification achieves 3.8% absolute improvement for micro-F1 measure. ©2016 Association for Computational Linguistics.

关键词： Linear transformations

来源：评论

学校读者我要写书评

暂无评论

Neuromorphic sampling on the SpiNNaker and parallella chip multiprocessors 7

Neuromorphic sampling on the SpiNNaker and parallella chip m...

引用

7th IEEE Latin American Symposium on Circuits and Systems, LASCAS 2016

作者： Mendat, Daniel R. Chin, Sang Furber, Steve Andreou, Andreas G. Department of Electrical and Computer Engineering Johns Hopkins University BaltimoreMD21218 United States Center for Language and Speech Processing Johns Hopkins University BaltimoreMD21218 United States Applied Physics Laboratory Johns Hopkins University LaurelMD20723 United States School of Computer Science University of Manchester ManchesterM13 9PL United Kingdom

ISBN: (纸本)9781467378352

We present a bio-inspired, hardware/software architecture to perform Markov Chain Monte Carlo sampling on probabilistic graphical models using energy aware hardware. We have developed algorithms and programming data flows for two recently developed multiprocessor architectures, the SpiNNaker and Parallella. We employ a neurally inspired sampling algorithm that abstracts the functionality of neurons in a biological network and exploits the neural dynamics to implement the sampling process. This algorithm maps nicely on the two hardware systems. Speedups as high as 1000 fold are achieved when performing inference using this approach, compared to algorithms running on traditional engineering workstations. © 2016 IEEE.

关键词： Markov processes

来源：评论

学校读者我要写书评

暂无评论

SHREC'16 Track: Partial Shape Queries for 3D object retrieval 9

SHREC'16 Track: Partial Shape Queries for 3D object retrieva...

引用

9th Eurographics Workshop on 3D Object Retrieval, 3DOR 2016

作者： Pratikakis, I. Savelonas, M.A. Arnaoutoglou, F. Ioannakis, G. Koutsoudis, A. Theoharis, T. Tran, M.-T. Nguyen, V.-T. Pham, V.-K. Nguyen, H.-D. Le, H.-A. Tran, B.-H. To, Q.H. Truong, M.-B. Phan, T.V. Nguyen, M.-D. Than, T.-A. Mac, K.-N.C. Do, M.N. Duong, A.-D. Furuya, T. Ohbuchi, R. Aono, M. Tashiro, S. Pickup, D. Sun, X. Rosin, P.L. Martin, R.R. Democritus University of Thrace Department of Electrical and Computer Engineering Greece ATHENA Research Center Institute for Language and Speech Processing Greece Norwegian University of Science and Technology Department of Computer and Information Science Norway University of Science VNU-HCM Viet Nam University of Science University of Information Technology VNU-HCM Viet Nam University of Science John von Neumann Institute VNU-HCM Viet Nam University of Illinois Urbana-Champaign United States University of Information Technology VNU-HCM Viet Nam Graduate School of Medicine and Engineering University of Yamanashi Japan Knowledge Data Engineering Lab. Toyohashi University of Technology Japan School of Computer Science and Informatics Cardiff University United Kingdom

ISBN: (纸本)9783038680048

Despite numerous recent efforts, 3D object retrieval based on partial shape queries remains a challenging problem, far from being solved. The problem can be defined as: given a partial view of a shape as query, retrieve all partially similar 3D models from a repository. The objective of this track is to evaluate the performance of partial 3D object retrieval methods, for partial shape queries of various scan qualities and degrees of partiality. The retrieval problem is often found in cultural heritage applications, for which partial scans of objects query a dataset of geometrically distinct classes. © 2016 The Eurographics Association.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Unsupervised speaker adaptation of deep neural network based on the combination of speaker codes and singular value decomposition for speech recognition 40

Unsupervised speaker adaptation of deep neural network based...

引用

40th IEEE International Conference on Acoustics, speech, and Signal processing, ICASSP 2015

作者： Xue, Shaofei Jiang, Hui Dai, Lirong Liu, Qingfeng National Engineering Laboratory of Speech and Language Information Processing University of Science and Technology of China Hefei China Department of Electrical Engineering and Computer Science York University Toronto Canada

ISBN: (纸本)9781467369978

Recently, we have proposed a general adaptation scheme for deep neural network based on discriminant condition codes and applied it to supervised speaker adaptation in speech recognition based on either frame-level cross-entropy or sequence-level maximum mutual information training criterion [1, 2, 3, 4]. In this case, each condition code is associated with one speaker in data, which is thus called speaker code for convenience. Our previous work has shown that speaker code based methods are quite effective in adapting DNNs even when only a very small amount of adaptation data is available. However, we have to use a large speaker code size and complex processes to obtain the best ASR performance since good initializations of speaker codes and connection weights are very important. In this paper, we propose a method using singular value decomposition (SVD) as in [5] to initialize speaker codes and connection weights to obtain a comparable ASR performance as before but with a smaller speaker code size and much less computation complexity. Meanwhile, we have evaluated unsupervised speaker adaptation with the proposed method in large vocabulary speech recognition in the Switchboard task. Experimental results have shown that it is effective for providing well initializations and suitable in adapting large DNN models. © 2015 IEEE.

关键词： Deep Neural Network (DNN) singular value decomposition (SVD) Speaker Adaptation Speaker Code

来源：评论

学校读者我要写书评

暂无评论

Robustness to additive noise of locally-normalized cepstral coefficients in speaker verification 16

Robustness to additive noise of locally-normalized cepstral ...

引用

16th Annual Conference of the International speech Communication Association, INTERspeech 2015

作者： Fredes, Josué Novoa, José Poblete, Victor King, Simon Stern, Richard Yoma, Nestor Becerra Speech Processing and Transmission Laboratory Electrical Engineering Department University of Chile Santiago Chile Institute of Acoustics Universidad Austral de Chile Valdivia Chile Centre for Speech Technology Research University of Edinburgh Edinburgh United Kingdom Department of Electrical and Computer Engineering and Language Technologies Institute Carnegie Mellon University PittsburghPA15206 United States

In this paper the performance of a new feature set, Locally Normalized Cepstral Coefficients (LNCC) is evaluated for a speaker verification task with short testing utterances in additive noise. The results presented here show that LNCC outperforms baseline MFCC features when SNR is lower than 15 dB. The average relative reduction in EER achieved by LNCC is 33%. The use of LNCC in combination with spectral subtraction provides a reduction in EER averaging 18% when compared to MFCC features also with spectral subtraction. In addition, sub-band LNCC is proposed to improve the estimation of noise energy and hence the effectiveness of spectral subtraction. When compared with MFCC features, the use of sub-band LNCC led to greater reductions in EER than LNCC with non-stationary noise. Copyright © 2015 ISCA.

关键词： Additive noise

来源：评论

学校读者我要写书评

暂无评论

Hybrid orthogonal projection and estimation (HOPE): a new framework to learn neural networks

The Journal of Machine Learning Research

引用

The Journal of Machine Learning Research 2016年第1期17卷

作者： Kevin Murphy Bernhard Schölkopf Shiliang Zhang Hui Jiang Lirong Dai Google MPI for Intelligent Systems National Engineering Laboratory for Speech and Language Information Processing University of Science and Technology of China Hefei Anhui China Department of Electrical Engineering and Computer Science York UniversityToronto OntarioCanada

In this paper, we propose a novel model for high-dimensional data, called the Hybrid Orthogonal Projection and Estimation (HOPE) model, which combines a linear orthogonal projection and a finite mixture model under a unified generative modeling framework. The HOPE model itself can be learned unsupervised from unlabelled data based on the maximum likelihood estimation as well as discriminatively from labelled data. More interestingly, we have shown the proposed HOPE models are closely related to neural networks (NNs) in a sense that each hidden layer can be reformulated as a HOPE model. As a result, the HOPE framework can be used as a novel tool to probe why and how NNs work, more importantly, to learn NNs in either supervised or unsupervised ways. In this work, we have investigated the HOPE framework to learn NNs for several standard tasks, including image recognition on MNIST and speech recognition on TIMIT. Experimental results have shown that the HOPE framework yields significant performance gains over the current state-of-the-art methods in various types of NN learning problems, including unsupervised feature learning, supervised or semi-supervised learning.

关键词： PCA mixture models neural networks orthogonal projection unsupervised learning von Mises-Fisher (vMF) model

来源：评论

学校读者我要写书评

暂无评论

Markov Chain Monte Carlo inference on graphical models using event-based processing on the SpiNNaker neuromorphic architecture

Markov Chain Monte Carlo inference on graphical models using...

引用

Annual Conference on Information Sciences and Systems (CISS)

作者： Daniel R. Mendat Sang Chin Steve Furber Andreas G. Andreou Department of Electrical and Computer Engineering Johns Hopkins University Baltimore Maryland Applied Physics Laboratory Johns Hopkins University Laurel Maryland School of Computer Science The University of Manchester UK Center for Language and Speech Processing Johns Hopkins University Baltimore Maryland

We present a combined hardware/software architecture to perform Markov Chain Monte Carlo sampling on probabilistic graphical models in a brain-inspired, energy-aware manner. By combining massively-parallel neuromorphic hardware architecture (SpiNNaker) with algorithms we've have developed for the event-based framework employed in SpiNNaker, we achieve large speedups when performing inference as compared to a traditional PC. We present results from two sampling approaches both well suited to the SpiNNaker architecture. Neural sampling, the first of the two approaches relies directly on simulating networks of spiking neurons while the second, Gibb's sampling is more flexible but still takes advantage of the hardware's event-handling capabilities.

关键词： Neurons Graphical models Markov processes Bayes methods Roads computer architecture MATLAB

来源：评论

学校读者我要写书评

暂无评论

Low-resource open vocabulary keyword search using point process models 15

Low-resource open vocabulary keyword search using point proc...

引用

15th Annual Conference of the International speech Communication Association: Celebrating the Diversity of Spoken languages, INTERspeech 2014

作者： Liu, Chunxi Jansen, Aren Chen, Guoguo Kintzley, Keith Trmal, Jan Khudanpur, Sanjeev Center for Language and Speech Processing Department of Electrical and Computer Engineering United States Human Language Technology Center of Excellence Johns Hopkins University BaltimoreMD United States

The point process model (PPM) for keyword search is a whole-word parametric modeling framework based on the timing of phonetic events rather than the evolution of frame-level phonetic likelihoods. Recent progress in PPM training and decoding algorithms has yielded state-of-the-art phonetic search performance in high-resource settings, both in terms of accuracy and computational efficiency. In this paper, we consider PPM application to low-resource settings where the amount of transcribed speech is severely limited and the pronunciation dictionary is incomplete. By using (i) state-of-the-art deep neural network acoustic models to generate phonetic events and (ii) grapheme-to-phoneme conversion to generate pronunciations for out-of-vocabulary (OOV) keywords, we find the PPM system reaches state-of-the-art OOV search performance at a small computational cost. Moreover, due to their complementary methodologies, combining PPM outputs with the LVCSR baseline produces average relative ATWV improvements of 7% and 50% for in-vocabulary and OOV keywords, respectively (16% overall). Copyright © 2014 ISCA.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Speaker adaptation of hybrid NN/HMM model for speech recognition based on singular value decomposition

Speaker adaptation of hybrid NN/HMM model for speech recogni...

引用

International Symposium on Chinese Spoken language processing

作者： Shaofei Xue Hui Jiang Lirong Dai National Engineering Laboratory of Speech and Language Information Processing University of Science and Technology of China Hefei Department of Electrical Engineering and Computer Science York University Toronto Canada

Recently several speaker adaptation methods have been proposed for deep neural network (DNN) in many large vocabulary continuous speech recognition (LVCSR) tasks. However, only a few methods rely on tuning the weight matrices in trained DNNs to optimize system performance since it is very prone to over-fitting especially when some class labels are missing in the adaptation data. In this paper, we propose a new speaker adaptation method for the hybrid NN/HMM speech recognition model based on singular value decomposition (SVD). We apply SVD on the weight matrices in trained DNNs, and then tune diagonal matrices with the adaptation data. This solves the over-fitting problem since we can change the weight matrices slightly by only modifying the singular values. We evaluate the proposed adaptation method in two standard speech recognition tasks, namely TIMIT phone recognition and large vocabulary speech recognition in the Switchboard task. Experimental results have shown that it is effective to adapt large DNN models using only a small amount of adaptation data. For example, the Switchboard results have shown that the proposed SVD-based adaptation method may achieve up to 3-6% relative error reduction using only a few dozens of adaptation utterances per speaker.

关键词： Hidden Markov models Adaptation models speech recognition Neural networks speech Matrix decomposition Signal processing

来源：评论

学校读者我要写书评

暂无评论

Discriminative confidence measure using linear combination of duration-based features and acoustic-based scores in keyword spotting

Discriminative confidence measure using linear combination o...

引用

International Symposium on Telecommunications, IST

作者： Mohammad Mohsen Goodarzi Yasser Shekofteh Iman Sarraf Rezaei Jahanshah Kabudian Speech and Natural Language Processing Group Research Center of Intelligent Signal Processing (RCISP) Tehran Iran Department of Computer Engineering and Information Technology Razi University Kermanshah Iran

One of the most important steps in a keyword spotting (KWS) system is a post-processing procedure to compute a confidence measure (CM) for each hypothesized keyword. The CM is commonly estimated by likelihood-based acoustic scores. However durations of the detected keyword, which include useful information, has not been studied directly in the KWS systems. In this paper, three duration-based features are proposed for such system. Also, using linear discriminant analysis (LDA), the proposed duration-based features are discriminatively combined with the acoustic scores and used as the final discriminative CM (DCM) in the KWS system. The proposed DCM results up to 11.3% improvement in FOM against to the conventional likelihood-based CM over a Persian conversational telephone speech database.

关键词： speech speech recognition Hidden Markov models Acoustics Acoustic measurements Databases Computational modeling

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：