Hierarchical Feature Reduction with Max Relevance and Low Dimensional Embedding Strategy and Its Application in Activity Recognition with Multi-sensors
Yu Wei, Libin Jiao, Rashid Mehmood, Liu Hua, Anton Umek, Anton Kos · Procedia Computer Science · 2018
Human activity recognition is widely discussed in many domains, and wearable sensors have proved to be a wise choice in related studies. Regarding activity estimation based on multiple sensors, we focus on the feature reduction process based on wearers’ experience and processing efficiency. According to the problem of determining the number and positions of necessary sensors in actual practice, we propose a hierarchical feature reduction method based on mutual information with max relevance and low-dimensional embedding strategies. This method divides the process of feature reduction into two stages: firstly, redundant sensors are eliminated with one-order sequential forward selection based on mutual information; secondly, feature selection strategy that maximizing class-relevance is integrated with low dimensional mapping so that the set of features will be further compressed.