INTERSPEECH 2008
9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Adaptive Filter Based Prosody Modification Approach

Qingcai Chen, Shusen Zhou, Dandan Wang, Xiaohong Yang

Harbin Institute of Technology, China

This paper proposes an adaptive filter based prosody modification approach. It consists of three components, i.e., a frequency estimator, a pitch tracker and a prosody modifier. Firstly, the frequency estimator is applied to segment the speech signal into sentences and the maximum frequency level of each sentence is estimated. Then, the adaptive filter that is adapted according to the maximum frequency level of each sentence is constructed. Its output is applied to estimate the pitch of speech. The prosody of the speech, which includes its pitch and duration, is then modified according to estimated pitch and the given pitch modification target. Be compared with existing techniques, the advantage of the proposed approach is its capability of adapting with the frequency of given speech. This advantage makes the approach especially suitable for the large range pitch modification. Finally, the performance of the proposed approach is evaluated and is compared with three existing prosody modification approach and the result is quit encouraged.

Full Paper

Bibliographic reference.  Chen, Qingcai / Zhou, Shusen / Wang, Dandan / Yang, Xiaohong (2008): "Adaptive filter based prosody modification approach", In INTERSPEECH-2008, 789-792.