基于组合参数的清浊音判决方法

The Method of Unvoiced/Voiced Decision of Speech Signal Based on Comprehensive Parameters

为了更简单地判决语音信号的清浊音,笔者提出了一种将自相关函数值分别与线性预测及偏相关系数相结合的判决方法,并阐述了各自的特征及相组合步骤;将该方法同过零率法相比较并进行了仿真研究,结果表明组合参数法的误判率低。

U/V decision is a important segment in speech processing. In general, the short-ZCR, short-energy and their combined method are used in U/V decision. Though these methods can achieve good results, it is necessary to improve the complexity. In order to get better, simpler and more advantageous method in this paper, we illustrate the algorithm that combining autocorrelation with LPC coefficients or PARCOR parameters to decide U/V. This algorithm,comparing with the ZCR method,has lower wrong decisive rate.