Traditional underwater sonar detection systems are primarily based on numerical methods such as pulse compression, Doppler velocity measurement, and beamforming to measure target distance, velocity, and azimuth parameters. In contrast, the sonar systems of organisms like bats rely on highly evolved neural perception to accomplish these tasks. By studying the detection mechanisms of biological sonar and developing bionic models, the target detection capabilities of underwater sonar systems can be enhanced. Inspired by Hipposideros Pratti., this paper designs a bat bio-sonar model for underwater target detection and provides theoretical derivations for various target parameters detection. A biomimetic sonar multi-harmonic signal waveform is designed based on multi-carrier modulation theory. Through the combination of different subcarrier components, the signal's penetration power is optimized, environmental noise interference is reduced, and target resolution and recognition accuracy are enhanced. The proposed waveform's excellent anti-reverberation performance is demonstrated through evaluations in underwater reverberation scenarios. For signal processing, this paper designs a parallel hierarchical processing architecture that can simultaneously handle different harmonic components sensing speed, distance, and azimuth information, significantly improving recognition accuracy and signal processing efficiency. To enhance the intelligence of bionic sonar systems, a parallel intelligent perception network model based on dilated convolution is proposed. It leverages feature maps of different harmonic groups to reduce the number of features required for extraction, improving the model's training efficiency and achieving intelligent perception of the sonar system. Simulation results indicate that the combination of different harmonic components can effectively perceive variations in target speed, distance, and direction, exhibiting strong anti-reverberation capability. Neural network recognition results show that the combination of different harmonics achieves an accuracy rate of over 95% for speed, distance, and azimuth recognition, verifying that the designed model has strong capabilities in underwater target perception.