Pitch Estimation by Multiple Octave Decoders

Complex-Domain Pitch Estimation Algorithm for Narrowband Speech Signals

“…Researchers have worked on pitch estimation for a long time [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34]. Because speech signals have pseudo-periodicity with a short frame length, time-domain pitch estimation algorithms estimate a pitch using an autocorrelation function [13], [14], [15], [16].…”

“…Kim et al employed a deep convolutional neural network to estimate the pitch, which directly operates on the time-domain signal [17]. Segal et al introduced a pitch estimation algorithm that deployed an encoder and multiple decoders to represent signal processing filterbanks [18]. Although their robustness can be enhanced by training predictive models from speech signals with noise, their applications are limited because retraining predictive models that is required for handling different sampling rates or frame lengths is time-consuming.…”