提出了一种基于LPC谱和离散小波变换的改进自相关基音频率检测算法,用于检测病理嗓音的基音频率。首先采用LPC谱特性滤除高频,得到预测余量,然后采用6级db20离散小波变换处理预测余量信号,并结合自相关法提取基音频率。采用美国MEEI公司的病理嗓音数据库进行基频检测实验。实验结果表明,本方法能够有效的去除倍频、半频的影响,是一种有效的病理嗓音基音频率检测算法。 An improved auto-correlation pitch detection algorithm based on LPC spectrum and discrete wavelet transform is proposed to detect the pitch frequency of pathological voice. Firstly, LPC spectral characteristics are used to filter out the high frequency, and the prediction residual is obtained. Then the 6-level db20 discrete wavelet transform is used to process the prediction residual signal, and the pitch frequency is extracted by combining the autocorrelation method. Using MEEI’s pathological voice database of the United States for the detection of fundamental frequency. Experimental results show that this method can effectively remove the influence of frequency doubling and half-frequency, and is an effective pathology voice pitch frequency detection algorithm.