Search for: All records

With the rapid growth of Machine Learning (ML) in recent years, more and more technical issues, which were usually solved by model-based solutions, have an opportunity to be solved with data driven solutions. Underwater Doppler effect was tackled with model-based solutions in tracking the motion and compensating the interference caused by multipath Doppler effect in communications. However, a too complex model for the harsh underwater conditions leads to massive computation and becomes an obstacle for the real-time Doppler compensation. In this research, we adopt ML techniques to solve underwater Doppler issues. We propose ML-based tracking and a tracking-aid ML-based compensation. The results show that joint tracking and compensation method have tap choosing accuracy 96.7%, 86.7%, 100% and 93.3% in different power ratios of the two-dominant path condition with fine tree, linear Support Vector Machine (SVM), quadratic SVM and cubic SVM.