Time Scale Modification of Speech Signals Using Wavelet Packet Transform

This paper presents methods for independently modifying the time and pitch scale of acoustic signals, with an emphasis on speech signals. The algorithm developed here is based on Wavelet Packet Transform followed by sinusoidal modeling of various sub - bands. A wavelet packet tree matching to the critical bands of human ear is designed and implemented. Sinusoidal model for the speech waveform is used to develop an analysis/synthesis technique that is characterized by the amplitudes, frequencies, and phases of the component sine waves. These parameters are estimated after decomposition of the speech signal into sixteen sub-bands using wavelet packet transform. The performance of the proposed algorithm is compared with the latest method available in the literature [1] which uses Short Time Fourier Transform and sinusoidal modeling. Simulation results show that proposed method gives improved quality in time and pitch scaling compared to the other method. Performance of the proposed algorithm is demonstrated using spectrum plots also.