基于线性源滤波器的语音频带扩展方法研究  

The Approach of Speech Bandwidth Expansion Based on Linear Source-Filter

作　　者：林胜义 肖政宏[1] 

机构地区：[1]广东技术师范学院计算机科学学院

出　　处：《自动化与信息工程》2016年第1期37-42,共6页Automation & Information Engineering

摘　　要：基于线性源滤波器模型的频带扩展方法,对矢量量化和隐马尔可夫模型这两种谱包络重建方法进行评测对比。实验结果表明:基于隐马尔可夫模型方法所恢复的宽带语音具有较低的失真度,其听觉质量要优于矢量量化方法恢复的语音。The wideband speech will become narrowband speech and have worse acoustic quality after transmitted through the communication networks.The bandwidth expansion is an effective way to recover the wideband speech from the narrowband speech.With the linear source-filter based bandwidth expansion implemented in this paper,authors make a comparison between Vector Quantization and Hidden Markov Model.The experiment results show that the wideband speech recovered by the Hidden Markov Model has less audio distortion and higher acoustic quality than the one recovered by Vector Quantization.

关 键 词：频带扩展 矢量量化 隐马尔可夫模型 

分 类 号：TN912.3[电子电信—通信与信息系统]