Fundamental Frequency Extraction of Noisy Speech Signals

Abstract

In this paper, we proposed a correlation based method which is a new approach using the autocorrelation function is weighted by the reciprocal of the YIN and very useful for accurate fundamental frequency extraction. The autocorrelation function and also YIN is a popular measurement in estimating fundamental frequency in time domain. In our proposed method, instead of the original signal, we employ its center clipping signal for obtaining the autocorrelation function and this function is weighted by the reciprocal of the YIN for fundamental frequency detection. Comparative results on female and male voices in white and exhibition noise shows that the proposed method can detect fundamental frequency with better accuracy in terms of gross pitch errors as compared to other related methods.