An Acoustic Signal Identification Method Based on Convolutional Neural Networks

Acoustic Signal Identification has become an important subject in the field of machine perception in recent years. It has achieved good results in application scenarios such as voice recognition, and it still has low precision in other Acoustic Signal recognition applications. Therefore, this paper proposes an acoustic signal recognition model based on convolutional neural network to improve the recognition accuracy. In this model, the first problem to be solved is the processing of acoustic source data. The model converts acoustic signals such as barking dogs, crying babies, waves and rain into one-dimensional spectral signals by using Fourier transform, and then inputs the data into one-dimensional CNN for training, and finally obtains the classification accuracy of ten categories of acoustic signals. The classification accuracy of this model CNN classifier is 69 %. In addition, this paper adds the pipeline micro-leakage data collected from actual engineering projects to the CNN model, and obtains better identification results. In general, this model outperform others.