Activity recognition by voting method via motion sensors

Classification of daily human movements using wearable sensor technology is among the current research topics, especially healthcare. In this study, eight different daily human movements (writing, walking, running, tooth brushing, writing on paper, using keyboard, stationary and cleaning) are classified by using data obtained from only smartwatch motion sensors (accelerometer and gyroscope). To compare the classification performances, machine learning methods was chosen from function based, rule based, tree based and Bayesian groups. Voting method is used as a meta level classifier. After the tests are performed, the most successful results are obtained from the voting method.