An experimental comparison between seven classification algorithms for activity recognition

Abstract

The daily activities recognition is one of the most important areas that attract the attention of researchers. Automatic classification of activities of daily living (ADL) can be used to promote healthier lifestyle, though it can be challenging when it comes to intellectual disability personals, the elderly, or children. Thus developing a technique to recognize activities with high quality is critical for such applications. In this work, seven algorithms are developed and evaluated for classification of everyday activities like climbing the stairs, drinking water, getting up from bed, pouring water, sitting down on a chair, standing up from a chair, and walking. Algorithms of concern are K-nearest Neighbor, Artificial Neural Network, and Naïve Bayes, Dynamic Time Warping, $1 recognizer, Support Vector Machine, and a novel classifier (D$1). We explore different algorithm activities with regard to recognizing everyday activities. We also present a technique based on $1 and DTW to enhance the recognition accuracy of ADL. Our result show that we can achieve up to 83 % accuracy for seven different activities.