在混合激励线性预测(MELP)语音编码算法中,语音帧的突变转换导致合成语音质量的下降成为一个突出问题.为解决该问题,提出一种基于过渡帧判决算法的改进MELP模型,提高了参数估计的准确度,有效实现了语音的降噪处理.-In the hybrid excitation linear prediction (MELP) speech coding algorithm, the abrupt change of the speech frame leads to the degradation of synthesized speech quality. The accuracy of the estimation, the effective realization of the voice of the noise reduction processing.