Investigation of different frequency estimation techniques using the phase vocoder

Sinusoidal models have long been used in the analysis and synthesis of audio signals. Many of these models including the vocoder choose critical frequencies in the amplitude spectrum for reconstruction in the synthesis section. In these models, it is important to estimate the true frequency of the signal. In this paper, we investigate the performance of traditional methods of frequency estimation using the phase vocoder. The performance of the frequency estimation techniques under varying signal to noise ratio (SNR) and different windows is analyzed. It will be shown that frequency estimation via the peak detection of the discrete Fourier transform (DFT) performs better at high SNR. We also study the effect of the presence of multiple sinusoids in the signal on the frequency estimation techniques. The mathematical analysis included is rigorous but has been presented in a simplified manner for easy interpretation