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 fr a me leads to the degradation of synthesized speech quality. The accuracy of the estimation, the effective realization of the voice of the noise reduction processing.