Hybrid continuous speech recognition systems by HMM, MLP and SVM: a comparative study

被引：21

作者：

Zarrouk, Elyes ^{[1
]}

Ben Ayed, Yassine ^{[1
]}

Gargouri, Faiez ^{[2
]}

机构：

[1] Natl Sch Engn Sfax, Sfax, Tunisia

[2] Sfax Univ, Higher Inst Comp Sci & Multimedia, Sfax, Tunisia

来源：

INTERNATIONAL JOURNAL OF SPEECH TECHNOLOGY | 2014年 / 17卷 / 03期

关键词：

Automatic speech recognition; Hybrid System; Hidden Markov Models; Multi layer perceptron; Support vector machines;

D O I：

10.1007/s10772-013-9221-5

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

This paper presents a new hybrid method for continuous Arabic speech recognition based on triphones modelling. To do this, we apply Support VectorsMachine (SVM) as an estimator of posterior probabilities within the Hidden Markov Models (HMM) standards. In this work, we describe a new approach of categorising Arabic vowels to long and short vowels to be applied on the labeling phase of speech signals. Using this new labeling method, we deduce that SVM/HMM hybrid model is more efficient then HMMs standards and the hybrid system Multi-Layer Perceptron (MLP) with HMM. The obtained results for the Arabic speech recognition system based on triphones are 64.68 % with HMMs, 72.39 % with MLP/HMM and 74.01 % for SVM/HMM hybrid model. The WER obtained for the recognition of continuous speech by the three systems proves the performance of SVM/HMM by obtaining the lowest average for 4 tested speakers 11.42 %.

引用

页码：223 / 233

页数：11

共 50 条

[1] A study on recognition of speech based on HMM/MLP hybrid network
Huang, XY
Ma, XH
Li, X
Fu, YQ
Lu, JR
[J]. 2000 5TH INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING PROCEEDINGS, VOLS I-III, 2000, : 718 - 721
[2] The hybrid ANN/HMM method with double MLP structure for continuous speech recognition
Lee, TZ
Chen, DW
[J]. PROGRESS IN CONNECTIONIST-BASED INFORMATION SYSTEMS, VOLS 1 AND 2, 1998, : 1096 - 1098
[3] An HMM/MLP hybrid approach for improving discrimination in speech recognition
Na, K
Chae, SI
[J]. IEEE WORLD CONGRESS ON COMPUTATIONAL INTELLIGENCE, 1998, : 156 - 159
[4] Hybrid SVM/HMM Model for the Recognition of Arabic Triphones-based Continuous Speech
Zarrouk, Elyes
Benayed, Yassine
Gargouri, Faiez
[J]. 2013 10TH INTERNATIONAL MULTI-CONFERENCE ON SYSTEMS, SIGNALS & DEVICES (SSD), 2013,
[5] Comparison and combination of features in a hybrid HMM/MLP and a HMM/GMM speech recognition system
Pujol, P
Pol, S
Nadeu, C
Hagen, A
Bourlard, H
[J]. IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, 2005, 13 (01): : 14 - 22
[6] Comparison between two hybrid HMM/MLP approaches in speech recognition
Fontaine, V
Ris, C
Leich, H
Vantieghem, J
Accaino, S
VanCompernolle, D
[J]. 1996 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, CONFERENCE PROCEEDINGS, VOLS 1-6, 1996, : 3362 - 3365
[7] A speech recognition system based on a hybrid HMM/SVM architecture
Qu Zhi-yi
Liu Yu
Zhang Li-hong
Shao Ming-xin
[J]. ICICIC 2006: FIRST INTERNATIONAL CONFERENCE ON INNOVATIVE COMPUTING, INFORMATION AND CONTROL, VOL 2, PROCEEDINGS, 2006, : 100 - +
[8] Chinese Speech Recognition Based on a Hybrid SVM and HMM Architecture
Luo, Xingxian
[J]. ADVANCES IN NEURAL NETWORKS - ISNN 2011, PT III, 2011, 6677 : 629 - 635
[9] A comparison between HMM and hybrid ANN-HMM based systems for continuous speech recognition
Ynoguti, CA
Morais, ED
Violaro, F
[J]. ITS '98 PROCEEDINGS - SBT/IEEE INTERNATIONAL TELECOMMUNICATIONS SYMPOSIUM, VOLS 1 AND 2, 1998, : 135 - 140
[10] HMM Topology in Continuous Speech Recognition Systems
Yared, Glauco F. G.
Violaro, Fabio
Selmini, Antonio Marcos
[J]. PROCEEDINGS OF THE IEEE INTERNATIONAL TELECOMMUNICATIONS SYMPOSIUM, VOLS 1 AND 2, 2006, : 651 - 656

← 1 2 3 4 5 →