A Comparison of Machine Learning Schemes for Moving Direction Estimation with Acoustic Data
In this paper, we investigate the problem of estimating the moving direction of a target using the acoustic data received by the horizontal array sonar in the shallow water environment. In this study, the two machine learning schemes of quadratic support vector machine and bagging tree are applied to this problem because the conventional methods showed poor performance in a rapidly changing environment like shallow water. To verify the learned model using quadratic support vector machine and bagging tree for a given environment, the performance of the learned models is compared and the correlation between the number of sensors and the performance of the models is experimentally investigated. As a result, it is observed that the model accuracy is proportional to the number of sensors and that the valid number of the sensor should be at least 13 and more. Furthermore, for all cases considered in the study, the bagging tree shows higher performance than the quadratic support vector machine.
horizontal array, passive SONAR,machine learning