码率可调节的高质量语音编码算法  

An adjustable bit rate high quality speech coder

作　　者：肖东[1,2] 莫福源[1] 陈庚[1] 马力[1] 

机构地区：[1]中国科学院声学研究所,北京100190 [2]中国科学院研究生院,北京100190

出　　处：《哈尔滨工程大学学报》2012年第8期956-965,共10页Journal of Harbin Engineering University

基　　金：国家自然科学基金资助项目(61102152)

摘　　要：中远距离水声语音通信时,由于水声信道可利用带宽窄的原因,通信速率较之无线电信道低.在高质量通话的前提下,语音编码码率也在一定范围内受到制约.在众多低速率语音编码标准中,美国联邦标准MELP(mixed excitation lin-ear prediction)是最佳选择,其编码码率2.4 kbit/s对于水声信道来说仍然偏高.考虑到水声信道的特点,结合语音信息不均匀分布、MELP编码参数分析和简化、码本冗余度降低等几个方面,从语音生成和听觉感知角度深入研究了MELP编码标准,采用不定帧数联合编码的方式,提出一种码率可调节的高保真语音编码算法.正常语速下,平均码率约800 bit/s.合成语音清晰可懂,保持了说话人的个性特征,其PESQ MOS(perceptual evaluation of speech quality mean opinion score)评分不低于2.7,语音质量接近2.4 kbit/s MELP标准水平,满足了中远距离(>10 km)水声高质量语音通信的要求.同时,本算法也可用于其他对实时性要求不高的场合.The channel bandwidth is limited in underwater acoustic communication.In medium or long distance underwater acoustic communication,the bit rate is required to be much lower than in a wireless communication channel.The speech coding bit rate is restricted to some extent,on the condition of high fidelity.The federal standard 2.4kbit/s speech coder,mixed excitation linear prediction（MELP）,is one of the best low bit rate speech coders,but its bit rate is still high for underwater acoustic communication.Considering the peculiarities of underwater acoustic channels,the asymmetrical distribution of speech information,MELP parameter analysis and refinement,and reduction of codebook redundancy,the MELP standard was studied from the viewpoints of speech synthesis and auditory perception.An adjustable bit rate high fidelity speech coder,whose number of united frames was not fixed,was proposed.At normal speech speed,the average bit rate was about 800 bit/s.The synthesized speech,which kept the individual characteristics of the speaker,was vivid and intelligible.The average perceptual evaluation of the speech quality mean opinion score（PESQ MOS） of synthesized speech was no less than 2.7,with a quality of close to 2.4 kbit/s MELP.The speech coder is available for medium or long distance（〉10 km） underwater acoustic high quality speech communication or other non-realtime situations.

关 键 词：语音编码 码率可调节 非实时 水声通信 

分 类 号：O423[理学—声学]