Autocorrelation analysis of speech signals using the number theoretic transforms over the direct sums of finite fields

In this paper we consider the application of the number theoretic transforms (NTTs) over the direct sums of finite fields R_m, m=0, 1, 2,... to autocorrelation analysis of a speech signal. We show that these NTTs are high-efficient and allow more flexible segmentation of the speech signal frame in comparison with the classical discrete Fourier transform (DFT). As a result, using the NTTs over R_m, m=0,1,2 reduces considerably the computational complexity of a speech signal frame autocorrelation function.