A comparison of classifiers for activity recognition using multiple accelerometer-based sensors

Physical activity has a positive impact on people's well-being and it can decrease the occurrence of chronic disease. To date, there has been a substantial amount of research studies, which focus on activity recognition using accelerometer and gyroscope-based sensors. However, many of these studies adopt a single sensor approach and focus on proposing novel features combined with complex classifiers to improve the overall recognition accuracy. In addition, the implementation of the advance feature extraction algorithms and the complex classifiers may exceed the computing ability of most current ambulatory monitoring sensor platforms. This study proposes a method to adopt multiple accelerometer-based sensors on different body locations to cope with this challenge. The objective of this method is to achieve higher recognition accuracy with 'light-weight' signal processing algorithms. For choosing the suitable classifier for this multi-sensor system, a comparison of the popular classifiers is presented with the same settings. Eight subjects were recruited to perform eight normal scripted activities in different life scenarios, and each repeated three times. Thus a total of 192 activities were recorded. These activities, then, were segmented and annotated in the laboratory. The collected dataset was used to compare and analyze the following classifiers: the Naïve Bayes classifier, the Decision Tree classifier, the Artificial Neural Networks classifier, the K-Nearest Neighbor classifier and The Support Vector Machines classifier. The comparison focuses on both recognition accuracy and execution time.