Signal reshaping using dominant harmonic for pitch estimation of noisy speech

A simple yet powerful method for pitch estimation of noisy speech is presented in this paper. The basic idea is to use the rectified dominant harmonic for signal reshaping. The main advantage of using the reshaped signal is the enhanced pitch peak as compared to other non-pitch peaks. The normalized autocorrelation function of the processed speech is used for pitch extraction. The performance of the proposed method is tested and compared with those of recent methods using the Keele pitch extraction reference database. Comparative results show that the proposed method can detect pitch with better accuracy in terms of gross and fine pitch errors as compared to other reported techniques.