R. Vergin et al., Generalized mel frequency cepstral coefficients for large-vocabulary speaker-independent continuous-speech recognition, IEEE SPEECH, 7(5), 1999, pp. 525-532
The focus of a continuous speech recognition process is to match an input s
ignal with a set of words or sentences according to some optimality criteri
a. The first step of this process is parameterization, whose major task is
data reduction by converting the input signal into parameters while preserv
ing virtually all of the speech signal information dealing with the text me
ssage. This contribution presents a detailed analysis of a widely used set
of parameters, the mel frequency cepstral coefficients (MFCC's), and sugges
ts a new parameterization approach taking into account the whole energy zon
e in the spectrum. Results obtained with the proposed new coefficients give
a confidence interval about their use in a large-vocabulary speaker-indepe
ndent continuous-speech recognition system.