WRIST-WORN ACCELEROMETER BASED ACTIVITY CLASSIFICATION USING DECISION TREE AND NEURAL NETWORK FOR SMART HEALTH APPLICATION

摘要

Automatic identification of human activity has led to a possibility of providing personalised services in different domains e.g. healthcare, security and sport etc. With advancement in sensor technology, automatic activity recognition can be done in an unobtrusive and non-intrusive way. The placement of the sensor and wearability are vital keys in successful activity recognition of free space living for smart health application. Experiments were carried out to investigate the use of a single wrist-worn accelerometer for automatic activity classification. The performances of two algorithms namely decision tree C4.5 and feed-forward backpropagation artificial neural network were compared on feature selection and activity classification processes. The results suggest that feature selection using ANN ranking with C4.5 classifier achieved the highest accuracy. The best accuracy of 94.17% was achieved using only a wrist-worn accelerometer showing a possibility of automatic activity classification with no movement constraint, discomfort or stigmatisation for continuous personalised smart health applications.