Abstract:

Aiming at the accuracy degradation problem caused by improper threshold setting for human fall detection threshold algorithms, we introduce the support vector machine to select thresholds. We obtain human motion information from an acceleration sensor, and extract acceleration and angle as classification features. We propose a fall detection model based on thresholds according to the three stages of human fall: free fall, hitting the ground and stationary. The fall detection model sets thresholds by the support vector machine and manual method respectively, and simulation results show that compared with the manual-threshold-setting method, the accuracy of the proposal is higher, which proves the effectiveness of the proposed method.

Key words: fall detection, threshold algorithm, support vector machine, acceleration sensor