A wavelet-based speech pitch detector for tone languages

The pitch of a speech signal is considered an important parameter in designing an automatic speech recognition system for tone based languages such as Chinese. Researchers have developed several different techniques for estimating the pitch of speech signals. One technique is an event-based pitch detection scheme utilizing the dyadic wavelet transform (DyWT). The present authors show how the DyWT can be used to detect the pitch in mono-tone Chinese speech. They have compared the results of the DyWT method with a standard pitch detection technique, modified autocorrelation, and obtained comparable and in some cases even better performance.<
>