This paper introduces a novel method for underwater acoustic target recognition (UATR) using the Residual Attentional Convolutional Neural Network (RACNN). The RACNN model integrates Mel Frequency Cepstral Coefficients (MFCC) feature extraction and deep learning techniques to enhance the recognition accuracy of ship-radiated noise. The MFCC features are extracted to mitigate noise interference, temporal variability, and frequency attenuation, while the RACNN model employs residual and attention mechanisms to improve learning capability, fault tolerance, and emphasis on critical information. The experimental results on the ShipsEar dataset demonstrate that the proposed model achieves an overall accuracy of 99.34%, surpassing conventional methods and other deep learning models. The RACNN model's performance is further validated through comparisons with VGG and ResNet, showing superior accuracy and efficiency. The study highlights the effectiveness of the RACNN model in recognizing underwater acoustic targets with high accuracy and reduced computational resources, making it a promising solution for maritime applications.This paper introduces a novel method for underwater acoustic target recognition (UATR) using the Residual Attentional Convolutional Neural Network (RACNN). The RACNN model integrates Mel Frequency Cepstral Coefficients (MFCC) feature extraction and deep learning techniques to enhance the recognition accuracy of ship-radiated noise. The MFCC features are extracted to mitigate noise interference, temporal variability, and frequency attenuation, while the RACNN model employs residual and attention mechanisms to improve learning capability, fault tolerance, and emphasis on critical information. The experimental results on the ShipsEar dataset demonstrate that the proposed model achieves an overall accuracy of 99.34%, surpassing conventional methods and other deep learning models. The RACNN model's performance is further validated through comparisons with VGG and ResNet, showing superior accuracy and efficiency. The study highlights the effectiveness of the RACNN model in recognizing underwater acoustic targets with high accuracy and reduced computational resources, making it a promising solution for maritime applications.