Walking recognition method for physical activity analysis system of child based on wearable accelerometer
Cheche Xie, Sheng Bi, Min Dong, Lan Li, Sunhuang Chi · 2017
Activities of daily living (ADL) recognition based on accelerometer is an increasingly important study, which is often treated as a multi-class classification problem and can be put into use widely in life. However, it treats all activities equally so that it can not obtain high accuracy in some specific activities, but also lacks specific research that combines application scenarios. This paper presents Hierarchical AdaBoost: a walking recognition method for physical activity analysis system of child based on wearable accelerometer. In order to improve the accuracy of classification, we improve AdaBoost classifier into two-layer classifier. The experimental results show that Hierarchical AdaBoost has higher performance comparing with one-class classification such as SVM, AutoEncoder, K-means and KNN for walking recognition.