PREPROCESSING AND CLUSTERING RAW ACCELEROMETER DATA FROM SMARTPHONES FOR HUMAN ACTIVITY RECOGNITION

In this paper is proposed an algorithm for preprocessing and clustering of raw data obtained by accelerometer sensor embedded into a smartphone. It is used by ordinary users while performing sitting, walking and running activities. The goal of the implementation is to enhance the representation of initially generated vectors into compact clusters. As the experimental results reveal it is necessary to introduce an advanced classification approach, such as SVM, in order to recognize the current activity. The method seems promising for application towards users with various medical conditions under remote and prolonged monitoring.

• Pavel Dinev
• Ivo Draganov
• Ognian Boumbarov

12th International Conference on Communications, Electromagnetics and Medical Applications (CEMA”17), 2017.

http://rcvt.tu-sofia.bg/CEMA2017_5.pdf