Pitch detection in time and frequency domain

Abstract

The beauty of human speech lies in the complexity of the different sounds that can be produced by a few tubes and muscles. This intricacy, however, makes speech processing a challenging task. One defining characteristic of speech is its pitch. Detecting this Pitch or equivalently, fundamental frequency detection of a speech signal is important in many speech applications. Pitch detectors are used in vocoders, speaker identification and verification systems and also as aids to the handicapped. Because of its importance many solutions to detect pitch has been proposed both in time and frequency domains. One such solution is pitch detection is by using Autocorrelation method and Average Magnitude Difference Function (AMDF), method which are analyses done in the time domain and the other is detecting the harmonic nature in the frequency domain. This paper gives the implementation results of the pitch period estimated in the time and frequency domains for vowel and fricative speech sounds, both for male and female speakers.