Pitch detection in time, frequency and cepstral domain for articulatory handicapped people

Abstract

Pitch is an important parameter of speech. A good estimation of pitch period is crucial to the performance of speech analysis and synthesis system. This paper describes autocorrelation and AMDF method for pitch detection in time domain. FFT, spectral autocorrelation method in frequency domain and cepstral domain algorithm for pitch detection. Voiced and unvoiced part of the speech is separated for further speech processing using pitch. Results are obtained for database of articulatoy handicapped people.