Robust pitch estimation with harmonics enhancement in noisy environments based on instantaneous frequency

The authors propose an approach for estimating pitch of speech in noisy environments based on instantaneous frequency (IF). First, they define the IF amplitude spectrum, which is obtained by projecting the STFT amplitude spectrum onto the IF axis. Based on the IF amplitude spectrum, one can perform harmonics enhancement by suppressing the aperiodic components. Next, they define an evaluation function to find pitch. This is done by expanding the IF amplitude spectrum to the time region. Then they propose a method for obtaining a continuous pitch contour using dynamic programming. Experiments show accuracy and robustness of the method especially when noise exists.