Pitch and formants extraction algorithm for speech processing
Adnane Chérif · 2000
We develop a new speech processing program which identifies the pitch period and extracts the formant frequencies of Arabic speech. Its purpose is to enhance voice recognition and to simplify complex processing. The database is composed of Arabic sentences phonetically balanced, pronounced by several speakers (male, female, children). After acquisition, conversion and segmentation, we identify the voiced/unvoiced (V/UV) speech by analysing the its zero-crossing evolution. Then we compute the fundamental frequency and the first four formants of the raw speech. The results are compared with other PDA such as SIFT, AMDF, CEPSTRAL, and LPC by computing the detection error and the V/UV decision error.