New Time-frequency Domain Pitch Estimation Methods for Speech Signals under Low Levels of Snr

New Time-Frequency Domain Pitch Estimation Methods for Speech Signals under Low Levels of SNR Celia Shahnaz, Ph.D. Concordia University, 2009 Pitch estimation of speech signals is the key to understanding most acoustical phenomena as well as accurately designing many practical systems in speech communication. It is to determine the fundamental frequency or period of a vocal cord vibration causing periodicity in the speech signal. This task becomes very difficult when the speech observations are heavily corrupted by noise. Although a large number of pitch estimation methods have been reported to deal with a noise-free environment, pitch estimation in the presence of noise has been attempted only by a few researchers. As noise generally obscures the periodic structure of the speech waveforms, many existing methods fail to provide accurate pitch estimates when the signal-to-noise ratio (SNR) is very low. The major objective of this research is to develop novel pitch estimation methods capable of handling speech signals in practical situations where only noise-corrupted speech observations are available. With this objective in mind, the estimation task is carried out in two different approaches. In the first approach, the noisy speech observations are directly employed to develop two new time-frequency domain pitch estimation methods. These methods are based on extracting a pitch-harmonic and finding the corresponding harmonic number required for pitch estimation. Considering that voiced speech is the output of a vocal tract system driven by a sequence of pulses separated by the pitch period, in the second approach, instead of using the noisy speech directly for pitch estimation,