Predictive Modeling of BPSK Bit Recovery Using Machine Learning in Underwater Acoustic Communication
2025 8th International Seminar on Research of Information Technology and Intelligent Systems, ISRITI 2025, Hybrid, Yogyakarta, Indonesia, 11 December 2025, pp.472-475, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/isriti68345.2025.11393377
- City: Hybrid, Yogyakarta
- Country: Indonesia
- Page Numbers: pp.472-475
- Keywords: BPSK, KNN, Logistic regression, Machine learning, Simple threshold, SVM, Thorp, UWA
- Middle East Technical University Northern Cyprus Campus Affiliated: Yes
Abstract
Underwater Acoustic communication UWA poses a significant challenge to researchers due to the difficulties associated with wireless signal propagation in the aquatic environment. The most significant reason leading to an increased error rate and reduced accuracy of the communication system is signal attenuation. Machine learning algorithms can find an algebraic formula between data inputs and outputs by using simulation data or experimental data. The more data is used, the more accurate the algebraic relationship becomes. Simple threshold, Logistic regression, Support Vector Machine SVM and K-Nearest Neighbors KNN were used to predict the received bit in the Binary Phase Shift Keying BPSK digital modulation scheme. The methods were applied using the simulation environment of the Thorp formula for the underwater acoustic channel. The simulation results showed that machine learning algorithms outperformed the thresholding method, achieving an accuracy of 99.84% when using SVM.