Patient's Motion Recognition Based on SOM-Decision Tree

摘要

Patient's motion recognition is quite popular in the area of healthcare and medical service nowadays. By analyzing the data from variant sensors within the network, we can estimate the activities a person does. The analyzing job is usually done by a classifier which can classify each motion into one category with similar movements. Self-Organizing Map (SOM) is a kind of algorithm that can be used to arrange data into different categories without any guidance. Decision tree is a mature tool for classification. In this paper, we propose a new kind of classification method with data from BAN called SOM-Decision Tree. Firstly, we use SOM on each of the sensor nodes to categorize motions into different classes, so that motions in different classes can be distinguished by this sensor. Secondly, a decision tree is constructed to discriminate each kind of movements from other motions. Finally, any action of the same patient can be recognized by query through the decision tree. According to our experiment, this algorithm is feasible and quite efficient.