Multi-rate HMM quantization for speech recognition

被引：0

作者：

Vasilache, Marcel ^{[1
]}

机构：

[1] Nokia, FIN-33721 Tampere, Finland

来源：

2008 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, VOLS 1-12 | 2008年

关键词：

hidden Markov models; quantization; speech recognition;

D O I：

10.1109/ICASSP.2008.4518616

中图分类号：

O42 [声学];

学科分类号：

070206 ; 082403 ;

摘要：

This paper refines the idea of scalar quantization for hidden Markov model (HMM) parameters which was introduced in an earlier contribution. With the proposed multi-rate approach it is shown that an increased model compression can be achieved with a significant computational complexity reduction while also closely preserving the recognition performance of the original models.

引用

页码：4341 / 4344

页数：4

共 9 条

[1] Subspace distribution clustering hidden Markov model [J].

Bocchieri, E ;

Mak, BKW .

IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, 2001, 9 (03) :264-275

[2]

KISS I, 2002, EUROSPEECH SCANDINAV, P1265

[3]

LEPPANEN J, 2005, INTERSPEECH EUROSPEE, P2965

[4] Direct training of subspace distribution clustering hidden Markov model [J].

Mak, BKW ;

Bocchieri, E .

IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, 2001, 9 (04) :378-387

[5]

Tan ZH, 2008, ADV PATTERN RECOGNIT, P1, DOI 10.1007/978-1-84800-143-5

[6]

VASILACHE M, 2004, ICASSP 2004, P113

[7]

VASILACHE M, 2000, ICSSLP 2000, P871

[8]

VASILACHE M, 2001, EUROSPEECH SCANDINAV, P1265

[9]

Viikki O, 1998, INT CONF ACOUST SPEE, P733, DOI 10.1109/ICASSP.1998.675369

← 1 →