Time-Frequency Domain Fundamental Frequency Estimation and Localization of Quasiperiodic, Pulsatic Signal in a Correlated Mixture

We propose a method for estimating the fundamental frequency and localizing a pulsatic, quasiperiodic signal in a correlated mixture. The spectra of the correlated signal is obtained by using the short-time Fourier transform (STFT). For a quasiperiodic, pulsatic signal, the frequency signal variation over the frame index is itself quasiperiodic and pulsatic. In a mixture, the overlap of the spectra of the second signal can erase this quasiperiodic and pulsatic property for a subset of frequency indices. Applying singular value decomposition (SVD) with different frame lengths to these frequency index signals, we can determine the length for which the sum of the ratio of the first over the second singular values is maximum. From this frame length, the fundamental frequency can be calculated. To localize the pulsatic signal, we apply the inverse STFT to the frequency bins where the ratio of singular values are large to obtain a time domain signal. This time domain signal gives an approximation to the position of the pulsatic signal in the mixture.