A pitch detector based on the dyadic wavelet transform and the autocorrelation function

Abstract

The interest in real-time, low-bit-rate speech coding dictates current research efforts in speech signal compression. This paper describes a pitch detection method based on the dyadic wavelet transform and the autocorrelation function (ACF). The proposed pitch detector ensures smooth pitch trajectory evolution. Experiments show the performance of the wavelet- and ACF-based pitch detector is superior to that of conventional pitch detectors that use the ACF method to estimate the pitch period.