Significance of phase in single frequency filtering outputs of speech signals

Studies on phase component of signals are important due to complementary information it provides besides the amplitude information. Though most studies focused on the phase of the short-time Fourier transform (STFT), there are other forms of phase like the phase of an analytic signal and the phase of the signals obtained through filtering operation on the signals. In this paper the significance of the phase of single frequency filtering (SFF) output of signals is examined. The single frequency filter has a pole on the negative real axis, close to the unit circle in the z-plane. The pole location parameter (r) controls the bandwidth of the filter at each frequency. Using sufficient number of filters in the SFF analysis, the speech signal can be reconstructed from the filtered signals without distortion. The relative importance of the SFF magnitude and SFF phase for reconstruction of the signal is examined for different values of r, and also by interchanging the magnitude and phase components obtained for two different values of r, as well as for two different utterances for the same value of r. It is observed that the intelligibility of the reconstructed signal is high if the SFF phase for values of r close to unity are used. The information in the reconstructed signal is dominated by the phase component of the SFF output.