检索结果-内蒙古大学图书馆

Conference on Advanced Signal Processing algorithms, Architectures, and Implementations IX

作者： Tung, TL Yao, K Reed, CW Hudson, RE Chen, D Chen, J Univ Calif Los Angeles Dept Elect Engn Los Angeles CA 90095 USA

ISBN: (纸本)0819432938

The problem of source localization from arrival time delay estimates requires a computationally costly iterative solution of a set of nonlinear equations. Most known methods assume that the propagation speed is known. In this paper, we provide several effective source localization and propagation velocity estimation methods which only use measurements of the relative arrival time delays between sensors. The formulae for source localization and propagation speed estimation are derived based on least squares, total least squares, bounded data uncertainty, and constrained least squares methods. Statistical performance of these methods are compared via computer simulation. In addition, in order to avoid time delay ambiguity problems and obtain smoother time delays, two time delay smoothing methods based on the forward backward algorithm and the Viterbi algorithm are also proposed. Field experiment results based on these techniques are also presented.

关键词： time delay smoothing source localization forward-backward algorithm Viterbi algorithm least squares total least squares bounded data uncertainty constrained least squares

来源：评论

学校读者我要写书评

暂无评论

SYNCHRONOUS TRANSFORMERS FOR END-TO-END SPEECH RECOGNITION

SYNCHRONOUS TRANSFORMERS FOR END-TO-END SPEECH RECOGNITION

引用

IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Tian, Zhengkun Yi, Jiangyan Bai, Ye Tao, Jianhua Zhang, Shuai Wen, Zhengqi Chinese Acad Sci Inst Automat Natl Lab Pattern Recognit Beijing Peoples R China Univ Chinese Acad Sci Sch Artificial Intelligence Beijing Peoples R China CAS Ctr Excellence Brain Sci & Intelligence Techn Beijing Peoples R China

ISBN: (纸本)9781509066315

For most of the attention-based sequence-to-sequence models, the decoder predicts the output sequence conditioned on the entire input sequence processed by the encoder. The asynchronous problem between the encoding and decoding makes these models difficult to be applied for online speech recognition. In this paper, we propose a model named synchronous transformer to address this problem, which can predict the output sequence chunk by chunk. Once a fixed-length chunk of the input sequence is processed by the encoder, the decoder begins to predict symbols immediately. During training, a forward-backward algorithm is introduced to optimize all the possible alignment paths. Our model is evaluated on a Mandarin dataset AISHELL-1. The experiments show that the synchronous transformer is able to perform encoding and decoding synchronously, and achieves a character error rate of 8.91% on the test set.

关键词： Asynchronous Problem Online Speech Recognition Synchronous Transformer Chunk by Chunk forward-backward algorithm

来源：评论

学校读者我要写书评

暂无评论

Joint Channel Estimation and Error Correction for Finite-State Markov Channels Using Polar Codes

Joint Channel Estimation and Error Correction for Finite-Sta...

引用

IEEE Global Conference on Communications (GLOBECOM) / Workshop on Wireless Networking and Control for UAV

作者： Ghaddar, Nadim Kim, Young-Han Milstein, Laurence B. Ma, Liangping Yi, Byung K. Univ Calif San Diego Elect & Comp Engn La Jolla CA 92093 USA InterDigital Commun LLC Wilmington DE USA

ISBN: (纸本)9781538647271

A joint channel estimation and channel coding scheme is presented for finite-state Markov channels using polar codes. Unlike the conventional approach of first estimating all channel parameters and then performing channel decoding separately, the proposed scheme incorporates a subset of reliable estimates of channel parameters into the decoding algorithm and computes decoding metrics averaged over the statistical behavior of the Markov channel. By adapting list decoding (without any inner code), channel estimation and decoding steps can be performed iteratively to boost the reliability of channel estimation as well as error correction. In order to improve the performance even further, a new pilot transmission scheme is developed that utilizes the structure of polar codes and sends pilot symbols along with code symbols. This construction can be viewed as a new family of shortened polar codes that can be of independent interest. Simulation results demonstrate the benefit of the proposed approach compared to existing solutions.

关键词： Polar codes channel estimation successive cancellation trellis decoding Viterbi algorithm forward-backward algorithm

来源：评论

学校读者我要写书评

暂无评论

On compositions of special cases of Lipschitz continuous operators

引用

FIXED POINT THEORY AND algorithmS FOR SCIENCES AND ENGINEERING 2021年第1期2021卷 25页

作者： Giselsson, Pontus Moursi, Walaa M. Lund Univ Dept Automat Control Lund Sweden Univ Waterloo Dept Combinator & Optimizat Waterloo ON N2L 3G1 Canada Mansoura Univ Fac Sci Dept Math Mansoura 35516 Egypt

Many iterative optimization algorithms involve compositions of special cases of Lipschitz continuous operators, namely firmly nonexpansive, averaged, and nonexpansive operators. The structure and properties of the compositions are of particular importance in the proofs of convergence of such algorithms. In this paper, we systematically study the compositions of further special cases of Lipschitz continuous operators. Applications of our results include compositions of scaled conically nonexpansive mappings, as well as the Douglas-Rachford and forward-backward operators, when applied to solve certain structured monotone inclusion and optimization problems. Several examples illustrate and tighten our conclusions.

关键词： Compositions of operators Conically nonexpansive operators Douglas-Rachford algorithm forward-backward algorithm Hypoconvex function Maximally monotone operator Proximal operator Resolvent

来源：评论

学校读者我要写书评

暂无评论

Multicomponent signal classification using the PMHT algorithm

Multicomponent signal classification using the PMHT algorith...

引用

5th International Conference on Information Fusion (FUSION 2002)

作者： Ainsleigh, P Luginbuhl, T USN Undersea Warfare Ctr Newport RI 02841 USA

ISBN: (纸本)0972184414

The probabilistic multi-hypothesis tracking (PMHT) algorithm is extended for application to classification. The PMHT model is reformulated as a bank of continuous-state hidden Harkov models, allowing for supervised learning of the class-conditional probability density models, and for likelihood evaluation of multicomponent signals.

关键词： classification probablistic multi-hypothesis tracking PMHT hidden Markov model HMM continuous-state HMM EM algorithm forward-backward algorithm

来源：评论

学校读者我要写书评

暂无评论

A predictive Hidden semi-Markov Model for bridges subject to chloride-induced deterioration 21

A predictive Hidden semi-Markov Model for bridges subject to...

引用

21st IEEE International Conference on Software Quality, Reliability and Security (QRS)

作者： Guo, Chunhui Liang, Zhenglin Zeng, Junqi Song, Minyuan Xue, Zongqi Tsinghua Univ Dept Ind Engn Beijing Peoples R China

ISBN: (纸本)9781665478366

Chloride-induced deterioration is one of the main deterioration mechanisms for bridges. Directly detecting the chloride ion concentration is uneconomical for most areas. Experienced workers estimate the chloride-induced deterioration through the corrosion of reinforced concrete, which results in ineffective maintenance due to the inaccurate estimation. Hence, we formulate a predictive Hidden semi-Markov Model for bridges subject to chloride-induced deterioration. The Hidden semi-Markov Models are capable of handling the "hidden states" scenario and have more general applications than the Hidden Markov Models. We derive the corresponding forward-backward algorithm of the Hidden semi-Markov Model to predict the future deterioration state based on the past observation sequences. A numerical example of reinforced concrete bridge decks is utilized to illustrate the applicability of the overall approach. In the numerical example, we discover that the predictive results of the Hidden semi-Markov Models outperform those of the Hidden Markov Models when we regard the deterioration trend from the physical model-Pick's second law as the benchmark. When the detection intervals are stochastic, the Hidden semi-Markov model is more practical than the Hidden Markov model.

关键词： chloride-induced deterioration Hidden semi-Markov Model forward-backward algorithm dectection interval

来源：评论

学校读者我要写书评

暂无评论

ENHANCED CONVERGENT PNP algorithmS FOR IMAGE RESTORATION

ENHANCED CONVERGENT PNP ALGORITHMS FOR IMAGE RESTORATION

引用

IEEE International Conference on Image Processing (ICIP)

作者： Terris, Matthieu Repetti, Audrey Pesquet, Jean-Christophe Wiaux, Yves Heriot Watt Univ Inst Sensors Signals & Syst Edinburgh EH14 4AS Midlothian Scotland Heriot Watt Univ Dept Actuarial Math & Stat Edinburgh EH14 4AS Midlothian Scotland Univ Paris Saclay INRIA Cent Supelec Ctr Visual Comp F-91190 Gif Sur Yvette France

ISBN: (纸本)9781665441155

Image restoration has long been one of the key research topics in image processing. Many mathematical approaches have been developed to solve this problem, e.g., variational methods, wavelet techniques, or Bayesian methods. With the widespread of neural network (NN) models in all the subdomains of data science, the performance limits of these methods are further pushed. One of the most successful strategies consists of plugging NNs in existing optimization algorithms. However, so doing raises several mathematical and practical challenges. One of the main issues is to secure the convergence of the resulting iterative scheme. Further questions concerning the characterization of the reached limit are also worth being addressed. In this paper, we show that the theory of maximally monotone operators allows us to bring insightful answers to these problems and to design firmly nonexpansive NNs;combining these with postprocessing NNs leads to excellent global restoration quality.

关键词： Neural networks forward-backward algorithm plug-and-play algorithm image restoration

来源：评论

学校读者我要写书评

暂无评论

Efficient minimum word error rate training of RNN-Transducer for end-to-end speech recognition 21

Efficient minimum word error rate training of RNN-Transducer...

引用

Interspeech Conference

作者： Guo, Jinxi Tiwari, Gautam Droppo, Jasha Van Segbroeck, Maarten Huang, Che-Wei Stolcke, Andreas Maas, Roland Amazon Com Seattle WA 98109 USA

ISBN: (纸本)9781713820697

In this work, we propose a novel and efficient minimum word error rate (MWER) training method for RNN-Transducer (RNN-T). Unlike previous work on this topic, which performs on-the-fly limited-size beam-search decoding and generates alignment scores for expected edit-distance computation, in our proposed method, we re-calculate and sum scores of all the possible alignments for each hypothesis in N-best lists. The hypothesis probability scores and back-propagated gradients are calculated efficiently using the forward-backward algorithm. Moreover, the proposed method allows us to decouple the decoding and training processes, and thus we can perform offline parallel-decoding and MWER training for each subset iteratively. Experimental results show that this proposed semi-on-the-fly method can speed up the on-the-fly method by 6 times and result in a similar WER improvement (3.6%) over a baseline RNN-T model. The proposed MWER training can also effectively reduce high-deletion errors (9.2% WER-reduction) introduced by RNN-T models when EOS is added for end-pointer. Further improvement can be achieved if we use a proposed RNN-T rescoring method to re-rank hypotheses and use external RNN-LM to perform additional rescoring. The best system achieves a 5% relative improvement on an English test-set of real far-field recordings and a 11.6% WER reduction on music-domain utterances.

关键词： end-to-end speech recognition RNN-Transducer discriminative sequence training minimum word error rate training forward-backward algorithm RNN-T rescoring LM rescoring

来源：评论

学校读者我要写书评

暂无评论

Leveraging Probabilistic Segmentation to Document Clustering 8

Leveraging Probabilistic Segmentation to Document Clustering

引用

8th International Conference on Contemporary Computing (IC3)

作者： Banerjee, Arko Coll Engn & Management Kolaghat India

ISBN: (纸本)9781467379489

In this paper a novel approach to document clustering has been introduced by defining a representative-based document similarity model that performs probabilistic segmentation of documents into chunks. The frequently occuring chunks that are considered as representatives of the document set, may represent phrases or stem of true words. The representative based document similarity model, containing a term-document matrix with respect to the representatives, is a compact representation of the vector space model that improves quality of document clustering over traditional methods.

关键词： document clustering boundary entropy forward-backward algorithm

来源：评论

学校读者我要写书评

暂无评论

EFFICIENT IMPLEMENTATION OF RECURRENT NEURAL NETWORK TRANSDUCER IN TENSORFLOW

EFFICIENT IMPLEMENTATION OF RECURRENT NEURAL NETWORK TRANSDU...

引用

IEEE Workshop on Spoken Language Technology (SLT)

作者： Bagby, Tom Rao, Kanishka Sim, Khe Chai Google 1600 Amphitheatre Pkwy Mountain View CA 94043 USA

ISBN: (纸本)9781538643341

Recurrent neural network transducer (RNN-T) has been successfully applied to automatic speech recognition to jointly learn the acoustic and language model components. The RNN-T loss and its gradient with respect to the softmax outputs can be computed efficiently using a forward-backward algorithm. In this paper, we present an efficient implementation of the RNN-T forward-backward and Viterbi algorithms using standard matrix operations. This allows us to easily implement the algorithm in TensorFlow by making use of the existing hardware-accelerated implementations of these operations. This work is based on a similar technique used in our previous work for computing the connectionist temporal classification and lattice-free maximum mutual information losses, where the forward and backward recursions are viewed as a bi-directional RNN whose states represent the forward and backward probabilities. Our benchmark results on graphic processing unit (GPU) and tensor processing unit (TPU) show that our implementation can achieve better throughput performance by increasing the batch size to maximize parallel computation. Furthermore, our implementation is about twice as fast on TPU compared to GPU for batch

关键词： recurrent neural network transducer forward-backward algorithm TensorFlow GPU TPU

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：