An ANFIS-based Human Activity Recognition using IMU sensor Fusion

Physical human activity is central to reducing the risk of many chronic diseases, thus it is considered vital to promoting healthy life styles. Also human Activity recognition (HAR) in recent times has found application in the explanations of the origin of complex diseases. In this paper an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based model is introduced for classification of daily living activities (ADLs) using data collected with a tri-axial Inertial Measurement Unit (IMU) sensor. Specifically, normalized data from the IMU axes within a specific window were considered to classify four chosen daily living activities (sit, stand, walk and run). The proposed ANFIS classifier were evaluated in terms of Root Mean Square Error (RMSE) over a variety of different ANFIS parameters and the results show that the selected activities are recognized well with an an overall accuracy rate of 98.88%.