Human Activity Recognition Using Wearable Accelerometer Sensors

Human Activity recognition has a wide range of applications such as remote patient monitoring, rehabilitation and assisting disables. Physical activity reduces the risk of many chronic diseases and is consider as a key factor for healthy life. In order to improve the state of global healthcare, numerous healthcare devices has been introduced that allows doctors to perform remote monitoring and increase users' motivation and awareness. Real time activity recognition systems encourage users to adopt healthier life style by increasing personal awareness about physical activities and its positive consequences on health. In this paper a machine learning based technique is proposed to enhance the accuracy of activity recognition system using feature selection method on an appropriate set of statistically derived features. A publically available HAR dataset on physical activities has been used in this work. Linear forward selection method is employed for feature selection. Activity classification is performed using Random forest and decision tree in connection with AdaBoost. The proposed technique outperforms the recent works.