Context-Aware Human Activity Recognition (CA-HAR) Using Smartphone Built-In Sensors

Context-awareness has the potential to enhance human activity recognition (HAR) by identifying daily activities such as driving, studying, cooking, or showering. Most existing context-aware HAR approaches that utilize smartphone sensors assume that the phone is placed on certain locations on the body such as trouser pockets, attached to the waist or arm, or held by hand. However, when the smartphone is no longer worn by the person, recognizing human activities becomes a challenging task. This paper proposes a context-aware human activity recognition (CA-HAR) approach to recognize human activities even when the smartphone is no longer placed on the body. The CA-HAR approach performs aggregation of multiple sensor data from the smartphone to recognize human activities by applying deep learning and ripple-down rules (RDR). It uses a context-activity model to build and formulate the RDR rules that consider additional contextual information to deal with the on-body location problem. The paper presents a proof-of-concept implementation of the CA-HAR as an Android app and discusses two types of evaluations that were conducted to validate the performance of CA-HAR.