Speech Prosody 2012

Shanghai, China
May 22-25, 2012

Prosody Modification for Vocoder Based on Amplitude Spectrum of Residual Signal

Zhengqi Wen, Jianhua Tao

National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, China

This paper describes the prosody modification (pitch and duration) for vocoder based on amplitude spectrum of residual signal. In this vocoder, period component is represented as amplitude spectrum of half pitch period length and aperiod component is estimated from the difference of amplitude spectrum between the constructed period signal and the residual signal. Then, pitch modification is conducted by resampling the period spectrum according to desired pitch period length in frequency domain and duration modification is conducted by adjusting the frame shift length in time domain. Listening tests show that the speech quality of proposed vocoder after modification is not decreased so much and can get comparable performance with STRAIGHT.

Index Terms: prosody modification, pitch, duration, amplitude spectrum

Full Paper

Bibliographic reference.  Wen, Zhengqi / Tao, Jianhua (2012): "Prosody modification for vocoder based on amplitude spectrum of residual signal", In SP-2012, 11-14.