ENHANCED DISTANCE MEASURE FOR LSP-BASED SPEECH RECOGNITION

被引：5

作者：

KIM, HK

KIM, KC

LEE, HS

机构：

[1] Department of Information and Communication Engineering, Korea Advanced Institute of Science and Technology, Dongdaemun-Gu, Seoul 130-012

来源：

ELECTRONICS LETTERS | 1993年 / 29卷 / 16期

关键词：

SPEECH RECOGNITION; SPECTRAL ANALYSIS;

D O I：

10.1049/el:19930979

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：

A distance measure based on the line spectrum pair (LSP) representation is proposed for improving the performance of speech recognition systems. The quefrency weighted cepstrum is to the root-power sum (RPS) of the inverse filter what the proposed measure is to the RPS of the filter whose roots are on the unit circle and expressed as LSP frequencies. The recognition results on a set of 10 confusable syllables show that the performance of this measure is better than that of the Euclidean distance measure for the LSP.

引用

页码：1463 / 1465

页数：3

共 5 条

[1] SPECTRAL SLOPE DISTANCE MEASURES WITH LINEAR PREDICTION ANALYSIS FOR WORD RECOGNITION IN NOISE [J].

HANSON, BA ;

WAKITA, H .

IEEE TRANSACTIONS ON ACOUSTICS SPEECH AND SIGNAL PROCESSING, 1987, 35 (07) :968-973

[2]

Oppenheim A. V., 1975, DIGITAL SIGNAL PROCE

[3]

PALIWAL KK, 1989, SPEECH COMMUN, V3, P27

[4]

SAITO S, 1985, FUNDAMENTALS SPEECH

[5] DYNAMIC-PROGRAMMING ALGORITHM OPTIMIZATION FOR SPOKEN WORD RECOGNITION [J].

SAKOE, H ;

CHIBA, S .

IEEE TRANSACTIONS ON ACOUSTICS SPEECH AND SIGNAL PROCESSING, 1978, 26 (01) :43-49

← 1 →