Activity Recognition using Actigraph Sensor

Numerous accelerometers are being extensively used in the recognition of simple ambulatory activities. Using wearable sensors for activity recognition is the latest topic of interest in smart home research. We use an Actigraph watch with an embedded accelerometer sensor to recognize real-life activities done in a home. Real-life activities include the set of Activities of Daily Living (ADL). ADLs are the crucial activities we perform everyday in our homes. Actigraph watches have been profusely used in sleep studies to determine the sleep/wake cycles and also the quality of sleep. In this paper, we investigate the possibility of using Actigraph watches to recognize activities. The data collected from an Actigraph watch was analyzed to predict ADLs (Activities of Daily Living). We apply machine learning algorithms to the Actigraph data to predict the ADLs. Also, a comparative study of activity prediction accuracy obtained from four machine learning algorithms is discussed.