Human Activity Recognition: A Step Toward Ambient Assisted Living

Human activity recognition (HAR) finds application from the field of ambient assisted living (AAL), health condition monitoring to surveillance, gaming, etc. Persuaded by the enormous application area, research activity in this field has also been escalated in last decade. Application of machine learning techniques for HAR is also famed. This work utilizes a very well-known machine learning technique random forest classifier to classify five activities of daily life (ADL). For sensing data, only one wrist worn tri-axial accelerometer is used, which makes this system cost effective and handy. To reduce time complexity of the classifier, dimensionality of training features needs to be reduced. Here to get this, dimension of the raw data has been reduced using principal component analysis. After that using a very few statistical and frequency domain features, the system offers 96% classification accuracy.