The judgment of unvoiced and voiced sound based on sparse representation in DCT dictionary is implemented. Human pronunciation can be mainly divided into unvoiced and voiced sound. Sparse representation can represent the signal with as few coefficients as possible on a set of over-complete vectors, which can reveal the most representative features of signals. In this paper, the difference between the sparse representation of unvoiced and voiced sound is studied, based on which a method is proposed to distinguish unvoiced and voiced sound in words. The experimental results prove that the proposed method is effective.
展开▼
