Spectral representations for speech recognition by neural networks-a tutorial
Biing Hwang Juang, L. R. Rabiner · 2003
Spectrum-based speech representations are discussed. Spectral representations, in order to be useful for speech recognition, need to be justified from both the computational (analytical) and the perceptual viewpoints. The authors' discussion of spectral representations, therefore, includes both the computational model and the associated measures of similarity that are appropriate for neural networks. This tutorial is intended to serve as a bridge between generic neural network classifiers and classical speech analysis for speech recognition applications. The various spectral representations discussed are intimately linked with appropriate spectral distortion measures that can be evaluated in the relevant domain of representation. The authors point out how these representations and spectral distortion measures can be applied in neural network solutions to pattern recognition problems.>