提升小波加权自相关函数的基音检测算法*

随着计算机技术的发展,语音信号处理作为人机交互的重要渠道,其在复杂噪声环境下的特征值检测算法直接关系到计算机的运算效率。基音周期是语音特征值提取的重要参数之一。针对传统基音检测算法在噪声环境下检测精度低的问题,提出了一种基于自适应提升小波变换加权线性预测误差自相关函数的基音检测算法。该方法用多级提升小波近似系数加权求和的方法来弥补自相关函数随着时间延迟量的增加幅值衰减的缺陷;用线性预测误差自相关函数的方法来抑制共振峰的干扰,然后将两种方法结合来突出基音周期处的峰值。实验结果表明,与传统的自相关函数法和小波加权法相比,该方法能有效减弱共振峰的影响,突出基音周期处的峰值,提高基音周期检测精度,鲁棒性更好。

Pitch detection based on lifting wavelet transform and weighted autocorreation

With the development of computer technology, speech recognition technology as an important channel of human-computer interaction, its eigenvalue detection algorithm is directly related to the computer"s computing efficiency in a complex noise environment. The pitch period is one of the important parameters of speech eigenvalue extraction. Aiming at the problem that the traditional pitch detection algorithm has low detection accuracy in noisy environment, a pitch detection algorithm based on lifting wavelet transform weighted linear predictive error autocorrelation function is proposed. This method use the way of Multi - level lifting wavelet approximation coefficient weighted summation to compensate the defect of the autocorrelation function decreases with the increase of the amount of time delay and the method of linear prediction error autocorrelation function is used to suppress the interference of the formant, then the two methods are combined with the peak at the pitch period. The experimental result shows that comparing with the traditional pitch detection algorithm, the method can effectively reduce the influence of the formant, highlight the peak at the pitch period, improve the accuracy of the pitch period detection and make robustness better.