Classification of Moving Ground Targets Using Accelerometer and Microphone Data

Abstract

In this study, an algorithm that can classify people, cars and empty environments was developed using microphone and 3-axis accelerometer data. k-Nearest neighbor classifier, Support Vector Machine and Decision Tree classifiers were used as classifier models. To solve the problem, sensor data was processed separately and classification was performed. Success analysis was made and it was observed that their performance was limited. In order to increase success, data augmentation and fusion methods have been proposed for classes with different sample numbers. Data preprocessing and feature extraction processes were applied. As a result of the experimental studies conducted with the classifiers used, the proposed data augmentation method and the fusion of two sensor data provided an increase in success in solving the problem.