A SVM Algorithm for Investigation of Tri-Accelerometer Based Falling Data
Thanh Hai Nguyen, Ty Phu Pham, Quoc Cuong Ngo, Thành Tâm Nguyên · American Journal of Signal Processing · 2016
Falling in elderly people is one of the main reasons causing serious injuries and increasing the risk of early death. Moreover, it can result in psychological problems from fear of falling. An automatic fall detection system is necessary for elderly people in daily activity alone. In this paper, a fall detection system applying a Support Vector Machine (SVM) algorithm is proposed for fall recognition. Data with different states collected by a trial-axis accelerometer system will be pre-processed using a mean filter for smoothing. In addition, features of the filtered signals will be extracted using a Principal Component Analysis (PCA). Therefore, the SVM will be employed to train feature data and then recognize fall states. Experiments will be performed many trials with eight different states on a subject and results will be processed to detect falling as well as to evaluate the accuracy of the proposed method.