Application of Cepstrum Algorithms for Speech Recognition

摘要

This paper proposes a novel method for speech recognition. This is method based on the calculation of correlation coefficients of cepestrum by linear prediction (LP). The results of the method is compared with other methods. The real cepstrum is used to separate/estimate the spectral content of the speech from its pitch frequencies. Some notes regarding the efficiency of the processors used in the simulations are also concluded.