论文部分内容阅读
在正弦类语音编码算法中 ,特征参数往往根据信号的浊清音判决结果由不同方式估计 ,相应地在解码端语音一般根据浊清音特性用不同算法合成 ,因此浊清音判决结果直接影响编码器的性能。本文设计了基于语音信号频谱特征的多带浊清音判决方法 ,该方法根据信号的谱平滑度、重建谱与原始谱的相似程度以及信号谱的相关性对语音信号分三步进行浊清音判决 ,有效地保证了浊清音判决的正确性。
In the sinusoidal speech coding algorithm, the characteristic parameters are often estimated by different ways according to the voiced unvoiced decision result of the signal. Correspondingly, the speech is generally synthesized by different algorithms according to the voiced unvoiced characteristic at the decoding side, so the voiced unvoiced decision result directly affects the performance of the encoder . In this paper, a multi-voiced unvoiced decision method based on the spectral features of speech signals is designed. According to the spectral smoothness of the signal, the similarity between the reconstructed spectrum and the original spectrum and the correlation of the signal spectrum, Effectively ensure the correctness of voiced unanimity.