摘要
多带激励声码器(MBE)由MIT的Grifin在1987年提出,其改进算法(IMBE)已被IN-MARSAT采纳作为卫星话音通信的标准。MBE声码器在中低速率上可获得较好的合成语音质量,但在噪声环境中使用时,随着输入信噪比的降低,其性能将显著恶化。本文试图将语音增强技术与MBE模型相结合以提高声码器抗噪声的性能。我们研究了两种方案:一是采用语音增强预处理器和IMBE声码器级联,二是将语音增强技术和IMBE声码器有机结合构成语音增强IMBE声码器。客观测试和主观试听表明,这两种系统在噪声环境中工作时,性能都有很大的提高。
The Multi Band Excitation vocoder was developed by Griffin in MIT in 1987. Subsequently the improved MBE was adopted by Inmarsat as a standard for satellite voice communications. It can yield high speech quality at medium and low rate speech coding, but its performance degrades dramatically in low SNR environments. This thesis intends to improve the performance of IMBE speech vocoder under noisy enviroment by the combination of speech enhancement techniques with the MBE speech model. So far two methods have been studied:IMBE vocoder with a speech enhancement pre processor and IMBE vocoder incorporated with speech enhancement technique. Objective measure and informal listening test indicate the improvement of both systems in noisy enviroment.
出处
《通信学报》
EI
CSCD
北大核心
1998年第4期43-49,共7页
Journal on Communications
关键词
多带激励模型
语音编码
语音增强
声码器
MBE model, DE model, speech coding, speech enhamcement, vocoder