Research on Feature Extraction Method for Acoustic Signals in Complex Systems with Multiple Excitations
In order to determine the operating status of the system,a new method combining Variational Modal Decomposition(IVMD)and Residual Network(ResNet)is used to identify different kinds of sound signals and deter-mine what kind of excitation the system is subjected to,so as to determine whether the system is working properly.First,the IVMD method was used to decompose the sound signal,and the center frequency ratio(CFR)was used as the evaluation index to determine the optimal K value of the variational modal decomposition(VMD);Then,the cor-relation coefficient(CC)and alignment entropy(AE)were combined to select three key eigenmode functions(IIMF)from the multiple eigenmode functions(IMF)obtained from the decomposition,and convert them into sound signal im-ages;Finally,the converted images were trained by taking advantage of the residual network in image processing.The experimental results show that the method achieves 99.57%accuracy for sound signal judgment and recognition,which is significantly better than other typical algorithms.
Improved variational modal algorithmsKey eigenmodal functionsResidual networksSound signal image
NETL
NSTL
万方数据
