Method of Representation of a Speech Signal in the Basis of Auditory Wavelets

In speech communication systems, the human auditory system acts as a recipient of speech messages, which makes correct decisions with great reliability about the presence of a message in the speech signal observed in noises. For a class of non-stationary signals, which includes a speech signal, their optimal representation is carried out in a basis built on the basis of wavelets. A new method of representing a speech signal in the basis of auditory wavelets consistent with it is proposed. Fast algorithms for direct and inverse wavelet speech transformation based on the auditory maternal wavelet have been obtained. On the example of processing a standard phonetically balanced speech phrase, the analysis of the proposed method is carried out. The effectiveness of the method is confirmed by the results of computer simulation. This result can be used to solve the problem of compact speech representation in low-speed voice codecs.