A Pitch Synchronous Method for Speech Modification

The speech modification is a mechanism of changing speech characteristics and prosody for some specific applications. It is used in voice conversion, pronunciation correction, tone perception, and language learning. The most important part is the change of pitch in an utterance. Pitch extraction is an essential process for speech modification. This paper presents an efficient pitch extraction algorithm based on the normalized second standard deviation function (NSSDF) of magnitude difference. A pitch synchronous method for modifying speaking rate and pitch trajectory is proposed. The speaking rate is modified by inserting or deleting pitch periods in voiced segments. The pitch trajectory change is accomplished by modifying the pitch period of residual signal obtained from pitch synchronous linear prediction (LP) analysis and reconstructing speech signal by LP filter. A speech modification system is developed for Mandarin perception which is used to help hearing impaired students in pronunciation learning.