ROOT CEPSTRAL ANALYSIS - A UNIFIED VIEW - APPLICATION TO SPEECH PROCESSING IN CAR NOISE ENVIRONMENTS

被引:45
作者
ALEXANDRE, P
LOCKWOOD, P
机构
[1] Matra Communication, 78392 Bois d'Arcy Cedex, Rue J.P. Timbaud
关键词
SPEECH RECOGNITION IN NOISE; ROOT CEPSTRAL ANALYSIS; NONLINEAR SPECTRAL SUBTRACTION (NSS); HIDDEN MARKOV MODEL (HMM); LINEAR PREDICTION;
D O I
10.1016/0167-6393(93)90099-7
中图分类号
O42 [声学];
学科分类号
070206 ; 082403 ;
摘要
The performance of speech recognition systems is significantly degraded in the presence of noise. To solve the noise problem, there is a need to reconsider standard approaches by taking into account this new constraint. We first envisage two well-known cepstral representations (parametric and non-parametric) of speech signals and propose a unifying view of both schemes. We introduce a pseudo-autocorrelation domain, which can be interpreted as a ''Root-cepstral domain'', and we show how non-parametric cepstral and linear predictive analyses converge to the same optimal solution. Experiments are carried out using an HMM-based isolated word recogniser for speaker-dependent and speaker-independent tasks in car noise environments.
引用
收藏
页码:277 / 288
页数:12
相关论文
共 22 条
[1]  
ALEXANDRE P, 1993, 1993 IEEE P INT C AC
[2]  
ATAL BS, 1974, J ACOST SOC AM, V55
[3]  
BATEMAN DC, 1992, MAR IEEE P INT C AC
[4]  
CHILDERS DG, 1977, OCT P IEEE, V65
[5]  
COHEN TJ, 1970, GEOPHYS J RAY ASTR S, V20
[6]  
DAVIS S, 1980, IEEE T ACOUST SPEECH, V28
[7]  
DURBIN J, 1960, REV I INT STATIST, V28
[8]  
FURUI S, 1986, IEEE T ACOUST SPEECH, V34
[9]   LOW-DIMENSIONAL REPRESENTATION OF VOWELS BASED ON ALL-POLE MODELING IN THE PSYCHOPHYSICAL DOMAIN [J].
HERMANSKY, H ;
HANSON, BA ;
WAKITA, H .
SPEECH COMMUNICATION, 1985, 4 (1-3) :181-187
[10]  
HIRSCH HG, 1991, 1991 P EUROSPEECH