Cer, the hydrophone receives the signalJ. Mar. Sci. Eng. 2021, 9,16 ofThe transmitted signal utilised within the offline education stage is composed of several information packets. The signal in each data packet is composed of an HFM signal and communication signal, in which the communication signal will be the known supply information and facts modulated by CSK-SS. Just after getting transmitted by the transducer, the hydrophone receives the signal just after passing via the water tank acoustic channel. Lastly, the received signal is processed because the coaching dataset on the neural network model to complete the education of your neural network model. Inside the on the net test stage, the hydrophone receives the transmitted signal and is directly input in to the trained neural network model after resampling to finish the signal demodulation. The evaluation from the simulation final results shows that inside the DL-based CSK-SS UWA communication system, the BiLSTM network has better efficiency than the LSTM network. Hence, the neural network model employed within the water tank experiment is BiLSTM plus the parameter settings in the neural network are consistent with these in the simulation. The key parameter setting and data structure with the transmitted signal are shown in Tables 2 and three.Table two. Parameter setting with the transmitted signal. Transmission Signal Parameters Spreading sequence Spreading sequence length Modulation mode Number of bits carried by every single symbol Chip length (ms) Sampling Rate (kHz) Center PF 05089771 Membrane Transporter/Ion Channel frequency (kHz) Table three. Parameter setting of the transmission signal frame structure. Transmission Signal Parameters in Every Frame Number of bits Communication signal duration (s) HFM signal duration (s) Gap duration (s) Communication rate (bps) Parameter Setting 120 0.93/0.62/0.465 0.1 0.1 106.19/146.34/180.45 Parameter Setting m-sequence 31 CSK-SS 2/3/4 0.five 484.2. Analysis of Experimental Results Within the water tank experiment, the functionality with the conventional method plus the DL-based system are compared. Furthermore, in CSK-SS modulation, 3 schemes with communication rates of 180.45 bps (each symbol carries 4-bit), 146.34 bps (every single symbol carries 3-bit), and 106.19 bps (every single symbol carries 2-bit) are analyzed. Also, in experimental data processing, noise interference is added artificially to get the situation of a low SNR. Inside the experiment, the coaching information samples with the neural network are received by hydrophones at positions B1, B2, B3, and B4, respectively. The data made use of to test the functionality with the DL-based method are received by the hydrophone at position C. As a consequence of the distinct places of the hydrophone, the UWA channel structures inside the training and test stage are also different. As a consequence of the change from the position of the getting hydrophone, the GS-626510 web unique CIRs is shown in Figure 16.J. Mar. Sci. Eng. 2021, 9,hydrophones at positions B1, B2, B3, and B4, respectively. The information utilized to test th formance in the DL-based program are received by the hydrophone at position C. D the various places from the hydrophone, the UWA channel structures within the trainin test stage are also different. Because of the change of your position of your receiving hydrop 17 of 22 the different CIRs is shown in Figure 16.Figure 16. Instruction stage andand stage: CIRs at various positions of hydrophone. Figure 16. Coaching stage test test stage: CIRs at diverse positions of hydrophone.The efficiency with the conventional and the DL-based preprocessing The overall performance on the traditional and the D.