清浊音分段判决的递推最小二乘自适应算法

Adaptive Voiced/Unvoiced Segmentation Based on RLS

在"零-能"判决法的基础上,结合递推最小二乘(RLS)对非平稳信号的自适应跟踪能力,提出自适应的清浊音分段算法.算法能够快速实现语音信号清浊音的精确分段,不需要通过样本集训练进行参数调整.其自适应能力是在单一话音样本上实现的,由RLS算法在清音段、浊音段及清浊音段交界处不同的跟踪能力来判别清/浊音段.与基于阈值的方法不同,算法基于极值点的识别,避免各种基于样本集训练的自适应学习算法在泛化能力上的缺陷,对于不同采样率、说话人、音量、背景噪声等变化因素,具有较强的自适应处理能力.

An adaptive voiced/unvoiced segmentation based on the traditional short-time analysis, with the adaptive tracking capacity of recursive least square （RLS） to the non-steady signal, has been presented. The algorithm can rapidly realize precise voiced/unvoiced segmentation, without parameter-adjustment by samples training. The adaptability comes from a single pronunciation sample, and deciding voiced/unvoiced segmentation based on the different tracking capacity of RLS in voiced/unvoiced section and the intersection point. It is different from the methods based on threshold, the algorithm based on recognizing the extreme value can avoid the drawback of various adaptive learning algorithms in generalization, which can better adapt for various varying factors of different sampling rate, speaker, volume, background noise.