Bispectral analysis of speech
Melvin J. Hinich, E. Shichor · 2002
Bispectral analysis is a more suitable signal processing approach for certain speech applications than conventional spectral analysis. There are two major reasons for applying the bispectral approach: (1) the bispectrum carries the signal phase information, and can thus be used to estimate the signal waveform and not just its spectrum; (2) since the third order moments of a Gaussian noise are all zero, the bispectral estimator is a better estimator of a voiced speech signal in the presence of Gaussian noise. Two examples demonstrate the advantages of this approach.>