Classifying human actions in daily life using computational intelligence techniques

Nowadays, there are several effective computational intelligence techniques that, theoretically, could be useful to classify human daily life actions. Moreover, sensors are getting smaller, cheaper, portable and even wearable. In this paper, we have built an annotation tool by applying several computational intelligence techniques (K-Nearest Neighbor, the Support Vector Machine and the Multilayer Perceptron) to detect six types of human actions in daily life based on signals obtained from an accelerometer sensor (standing-up, walking, running, resting, jumping and sitting-down) with an accuracy over 85%. In the future, this component will be the base to infer abnormal behavior from common daily behavior that could be an emergency situation in evolution.