音频编码中瞬态信号的时域检测方法Time-Domain - 电子与信息学报.PDFVIP

第28 卷第2 期 电 子 与 信 息 学 报 Vol.28No.2 2006 年2 月 Journal of Electronics Information Technology Feb. 2006 音频编码中瞬态信号的时域检测方法 阎建新 窦维蓓 董在望 (清华大学电子工程系 北京 100084) 摘 要 在低比特率音频感觉编码中，预回声失真更为突出，而对其处理的前提是瞬态信号的有效检测。在时域， 基于峰值功率与平均功率之比( PMR)定义了瞬态强度，并以此为判决函数提出一种新的瞬态信号时域检测算法。 由于考虑了时域掩蔽效应来设置检测门限和有效瞬态点间隔，非常适用于感觉音频编码。与当前典型的基于感觉 熵的频域瞬态检测方法相比，具有时间分辨率高、准确和算法简单等优点。 关键词 语音处理，瞬态检测，预回声失真，音频感觉编码 中图分类号：TN912.3 文献标识码： A 文章编号：1009-5896(2006)02-0307-05 Time-Domain Detection Method of Transient Signals in Audio Coding Yan Jian-xin Dou Wei-bei Dong Zai-wang (Dept. of Electronic Eng., Tsinghua University, Beijing 100084, China) Abstract In perceptual audio coding with low bit rate application, the pre-echo artifact is more distinct, and the effective detection of transient signals is the precondition. Transient intensity is defined in term of the Peak-Mean power Ratio(PMR), accordingly, a novel detecting algorithm of transient signals on time domain is presented while PMR is used as a criterion function. Because of considering the temporal masking effects to set up a threshold and the valid hop size of transient points, this method is very suitable for perceptual audio coding. When compared with the typical perceptual entropy detecting on frequency domain，it has some advantages, such as higher time resolution, more accuracy and simple algorithm. Key words Speech processing, Transient detection, Pre-echo artifacts, Perceptual audio coding 1 引言 频域进行量化编码时，由于量化噪声被扩散到整个变换块 (或滤波器组)范围上，而且如果不能被信号掩蔽，就会出现 感觉音频编码的一般过程是：(1)使用滤波