Underwater Acoustical Sensing for Target Localization With Rational Orthogonal Wavelet Signaling

Underwater moving target localization has long been a challenging research task due to multipath propagation, the Doppler effect, and underwater acoustical ambient noise. The resultant severe signal distortions contain rich features that can be utilized for target sensing. This article proposes the transceiver design of a broadband active sonar system, which explores the unique features of multipath and Doppler. Six waveforms, i.e., continuous wave signals, linear frequency modulated signals, and four rational orthogonal wavelet (ROW) signals with different scaling factors, are selected as the transmit signals. In the receiver design, a new feature extraction method, ROW frequency cepstral coefficient, is proposed and evaluated with other feature extraction methods, including short-time Fourier transform, gammatone frequency cepstral coefficient, linear prediction cepstral coefficients, and perceptual linear prediction coefficients. A lightweight CNN model is adopted as the classifier following the feature extraction. The receiver design is further improved with the sensor array. A large database is generated to include the received echo signals with different parameter settings that match specific target sensing scenarios. Comprehensive simulations are conducted to demonstrate the effectiveness and superiority of the proposed wavelet-based target localization system in handling multipath, Doppler, and low signal-to-noise ratios.