7th International Conference on Spoken Language Processing

September 16-20, 2002
Denver, Colorado, USA

Warped-LP Residual Resampling Using DCT for Pitch Modification

R. Muralishankar, A. G. Ramakrishnan, P. Prathibha

Indian Institute of Science, India

In this paper, we propose a novel algorithm for pitch modification. The linear prediction (LP) residual is obtained from pitch synchronous frames by inverse filtering the speech signal. Then the Discrete Cosine Transform (DCT) of these residual frames is taken. Based on the desired factor of pitch modification, the dimension of the DCT of the residual is modified by truncating or zero padding, and then the Inverse DCT is obtained. This period modified residual signal is then forward filtered to obtain the pitch modified speech. The mismatch in the positions of the harmonics between the pitch modified signal and the LP spectrum introduce gain variations, which is more pronounced in the case of female speech [2]. This is minimised by modifying the radii of the poles of the filter to smoothen the peaky linear predictive spectrum before forward filtering. Since Warped Linear Prediction (WLP) [7] exhibits frequency resolution close to human hearing, it has been exploited for our pitch modification algorithm. WLP coefficients are used instead of the conventional LP coefficients for pitch modification. Perceptual results show better performance of WLP over conventional LP. The technique has been successfully applied to create interrogative sentences from affirmative sentences.

Full Paper

Bibliographic reference.  Muralishankar, R. / Ramakrishnan, A. G. / Prathibha, P. (2002): "Warped-LP residual resampling using DCT for pitch modification", In ICSLP-2002, 2517-2520.