论文部分内容阅读
针对恶劣移动音频传输环境下突发连续大量丢帧问题,本文提出一种基于HMM的丢帧隐藏方法,通过分析语音信号在更大范围的上下文关系的统计学变化来选择合适的丢帧隐藏策略。当包丢失时,基于HMM的恢复方法使用状态和密度函数信息,计算丢失帧参数的估计值。实验结果表明:提出的方法相比AVS-P10标准的语音编码器原有方法,客观语音测试PESQ平均分提高约0.33分,主观语音测试CMOS平均分能够提高约0.05分。
Aiming at the problem of a large number of dropped frames in a burst under a bad mobile audio transmission environment, this paper proposes a HMM-based frame loss concealment method, which selects the appropriate frame loss concealment strategy by analyzing the statistical changes of speech signals over a wider range of contexts . When a packet is lost, the HMM-based recovery method uses the state and density function information to calculate an estimate of the missing frame parameters. The experimental results show that compared with the original AVS-P10 speech coder, the proposed method improves the average PESQ score of objective speech test by about 0.33 and the average score of subjective speech test CMOS by about 0.05.