基于替代数据检测的音频信号非线性分析

Nonlinear Analysis of Audio Signals Using Surrogate Data Test

针对轻微含噪的音频信号,本文提出了一种基于替代数据检测的非线性分析方法。该方法首先假设音频信号来自于线性高斯随机过程,并利用基于随机相位的傅里叶变换法生成多组替代数据,然后分别计算原始数据和替代数据的样本峰度,最终根据假设检验方法判断原始音频中是否包含非线性成分。实验测试中分别对不同信噪比下音调型器乐演奏信号进行了分析。结果表明,相比于传统基于最大Lyapunov指数的非线性分析方法,本文所提方法在含噪情况下能更准确地检测到音频信号的非线性成分。

A kind of nonlinear analysis method based on the surrogate data test for slightly nois- y audio signals is presented. According to the hypothesis that audio signals can be modeled as a linear Gaussian stochastic process, several groups of surrogate data are generated by means of Fourier transform with random phase. The kurtosis measures are respectively calculated for the original audio and surrogate data. Hypothesis test is utilized to detect the nonlinear compo- nents in the original audio. In experiments, the audio signals of harmonic instruments were an- alyzed in the condition of different signal-to-noise ratios. The results show that the proposed method achieves the better performance for detecting the nonlinear components of audio signalsthan the nonlinear analysis based on the largest Lyapunov exponent in the noisy conditions.