基于LPC和基音周期的方言语音合成改进算法OA

Aiming at the serious loss of vocal parameter information and poor synthesis effect in Ankang dialect speech synthe-sis using linear prediction coefficient(LPC),an improved speech synthesis algorithm combining LPC and pitch period is proposed.On the basis of the traditional algorithm,the improved agent-extended pitch detection algorithm is used to extract the pitch parame-ters of the speech signal.Then the calculation accuracy is improved by gating and amplifying the excitation source model.Finally,through the filter model controlled by excitation pulse sequence parameters and resonance peak parameters,the synthesized Ankang dialect speech signal is better obtained.According to the experimental results,the improved algorithm in this paper can reduce the root mean square error(RMSE)by 7.14%compared with the traditional algorithm,and has an ideal synthesis effect.The synthe-sized pitch period and voice waveform are observed and analyzed.The improved algorithm in this paper can preserve more content and details of Ankang dialect voice signal.