基于VMD的声音信号增强算法研究

Algorithm of Sound Signal Enhancement Based on VMD

为提高声音信号增强效果,减小实际信号的计算量,利用可变模态分解(Variational mode decomposition,VMD)与相关系数以及能量的起始点检测准则相结合提出一种新的信号增强算法。该算法首先利用能量的起始点检测准则判断出实际信号的起始点提取有效信号;利用VMD分解该信号,计算分解后各本征模态函数(Intrinsic mode function,IMF)与原始信号的相关系数;利用预设的相关系数阈值来自适应确定有效IMF,利用有效的IMF重构信号。为了评估该算法的增强效果,利用该算法与经验模态分解(Empirical mode decomposition,EMD)算法进行对比分析。理论分析和试验结果表明:提出的算法在相同信噪比不同采样频率以及不同输入信噪比的条件下获得的输出信噪比都高于EMD算法,从而验证了该算法的稳定性和准确性。

In order to improve the signal enhancement effect and reduce the calculation of the actual signal, a new signal enhancement algorithm is proposed by combining the variational mode decomposition(VMD) with the correlation coefficient and the starting point detection criterion of energy. The algorithm uses the starting point detection criterion of energy to judge the starting point of the actual signal to extract the effective signal. Then, the VMD is used to decompose the signal to calculate the correlation coefficient of the intrinsic mode function(IMF) and the effective signal. It extracts useful IMFs by defaulting the correlation coefficient threshold, and reconstructs the filtering signal using the useful IMFs. In order to evaluate the enhancement effect of the algorithm, the algorithm is compared with the empirical mode decomposition(EMD) algorithm. The theoretical analysis and experimental results show that the proposed algorithm has higher output signal to noise ratio than the EMD algorithm under different sampling frequency and different input signal-to-noise ratio, which verifies the stability and accuracy of the algorithm.