【作者】 吴超； 杨震；

【Author】 Wu Chao,Yang Zhen (Institute of Signal and Information Processing;Nanjing University of Posts and Telecommunications; Nanjing China;210003)

【机构】 南京邮电大学信号与信息处理研究所；

【摘要】 宽带语音(50-7000Hz)为语音实时通信提供了更高的话音质量。本文设计了一种可以在大多数现有的窄带语音编码算法中嵌入使用的宽带语音编码算法,采用分带方法,算法用较低的码率(700kbit/s)单独编码并传输宽带语音的高频部分,而窄带部分采用现有的语音编码算法。对高频部分使用正弦模型建模,分析提取谱包络参数,解码器利用了窄带码流的部分解码参数参与高频信号的正弦合成。本文以8kbit/s代数码激励线性预测(ACELP)编码器为例处理窄带语音,在8.7kbit/s的码率上实现了对宽带语音的编解码,非正式的主观听觉测试表明本文算法的解码语音略优于AMR-WB(G.722.2)在8.85kbit/s模式时的性能,而且可以在大多数现有的窄带语音编码算法中嵌入使用,有一定的实用价值和可操作性。更多还原

【Abstract】 Wideband speech coders are capable of achieving higher quality for real-time speech communications.This paper designs a wideband speech coder which can be embedded into most of the existing narrow band speech coding algorithms.The algorithm uses split-band method first,the high-band signal is coded and transferred independently using comparatively lower bit rate(700bit/s),and the narrow band signal is processed by an existing speech coding algorithm.The high-band signal is analyzed using a sinusoidal model to extract high-band spectral envelope parameters.The high-band decoder synthesize high-band signal using some parameters from narrow band decoder.We take the 8kbit/s ACELP codec for example to process narrow band speech, and achieve to code the wideband speech at the bit rate of 8.7kbit/s.Our algorithm performs better than the 8.8Skbit/s mode of AMR-WB(G722.2) in the informal subjective tests,and this algorithm can also be embedded into most of the existing narrow band system.It has some practicality and maneuverability to a certain extent.更多还原