Ship Radiation Noise Identification Method Based on CNN

Ship radiated noise recognition has important application value in the field of underwater acoustic target recognition, but existing methods face problems such as insufficient feature extraction, low classification accuracy, and weak anti-noise interference ability. In response to these challenges, this article proposes a ship radiated noise recognition method based on Convolutional Neural Network (CNN). Firstly, this article preprocesses the collected ship radiation noise signals, including noise reduction, normalization, and frame segmentation; Secondly, this article converts the time-domain signal into a time-frequency graph and uses Short Time Fourier Transform (STFT) to extract the time-frequency characteristics of the signal; Then, this article constructs a multi-layer CNN model, through convolution operations, it automatically learns deep feature representations of noise signals, and finally uses a Softmax classifier to achieve ship type recognition. The experiment used the ShipsEar dataset, which includes 1000 samples of 5 types of ships. The results showed that the recognition accuracy of proposed method reached its peak at 94.3% in the 60th round, an increase of 12.5 percentage points compared to SVM’s 81.8% and an increase of 8.7 percentage points compared to BPNN’s 85.6%. It still maintained a recognition accuracy of 85.5% under the condition of a signal-to-noise ratio of 5dB.