基于听觉外围模型的音频基频估计方法

Fundamental frequency estimation method for audio signals based on auditory peripheral model

针对音频信号中的暂态成分对基频检测可靠性的影响,提出了一种基于听觉外围模型的基频估计方法。该方法根据听觉外围模型来模拟声音在人耳内耳听觉神经上传导过程,并利用循环平均幅度差函数判断每个神经传导信号上呈现的时域周期性,进而提取音频信号的基频。实验结果表明,在纯净音频条件下,该方法能准确地估计出音频信号的基频,并且在不同音量打击乐信号的干扰下,所提方法的平均错误率低于三种参考方法。

The transient components of audio signals may reduce the reliability of fundamental frequency estimation. A fundamental frequency estimation method based on auditory peripheral model is proposed in this paper. The auditory peripheral model is adopted to mimic the transduction in the auditory nerve of the inner hair cells, and a circular average magnitude difference function is used to judge the periodicity of the response of auditory nerve for estimating the funda-mental frequency of audio signals. Experimental evaluations show that the proposed method can accurately estimate the fundamental frequency of the clean audio signals. With the percussive interference in different volumes, the proposed method also achieves a lower error rate in comparison with three reference methods.