Underwater target classification using a deep neural network model based on fast Fourier transform

Detection and classification often involve converting visual images or audio data into images, which are then input into a Convolutional Neural Network (CNN). However, this approach requires significant computational resources, necessitating high-performance hardware and imposing a substantial burden on the system, particularly for real-time detection. This study proposes the use of a Deep Neural Network (DNN) model for real-time underwater target classification through frequency analysis. Preprocessing is necessary to apply audio data to the DNN model. In this study, audio data are transformed into the frequency domain using Fast Fourier Transform (FFT). The values of each frequency domain are separated by commas, and the class is stored at the end of each row to create a text file dataset. By training the model on frequency ranges that are characteristic of each class, unnecessary data can be excluded, allowing for the use of a relatively simple deep-learning model. Based on experimental results, the performance of the proposed DNN model is verified by comparing it with existing CNN models. The DNN model proposed in this study is expected to enable underwater target classification with fast processing and low costs.