Transform based pitch-scale modification algorithms for speech

• Main Authors: Yu-Hao Chang, 張育豪
• Other Authors: Fu-Kun Chen
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/61218956144197743637

碩士 === 南台科技大學 === 資訊工程系 === 99 === Pitch scale modification (PSM) has been widely used in language learning, kara ok, Text-to-Speech (TTS), entertainment and digital learning. This thesis proposed three PSM algorithms: PSM based on DCT (DCT-PSM), SOLA based linear prediction PPSM (SOLA-LPDCT) and Modified SOLA-LPDCT. The proposed DCT-PSM algorithm uses discrete cosine transform (DCT) instead of discrete Fourier transform (DFT) to rearrange the spectrum for pitch shifting. The SOLA-LPDCT algorithm uses the time-scale modification algorithm to capture the suitable frame for PSM processing. And then, the linear prediction coding (LPC) is performed for PSM processing in the LPC residual signal domain. Finally, the modified SOLA-LPDCT approach tries to improve the performance by using the first order interpolation in the PSM processing of SOLA-LPDCT method. Simulation results show that the proposed algorithms not only reduce the computational complexity but also improve the performance while comparing the mentioned methods in literature.