Fall Detection Method Based on TBM and SVM

In response to the low power consumption and high accuracy requirements of wearable fall detection technology, a fall detection method based on threshold-based method (TBM) and Support Vector Machine (SVM) was proposed in this paper. TBM is implemented in the designed wearable device, and SVM is implemented in the server. In the TBM, three features corresponding to the weightless, impact and stationary of falls were extracted. They were compared with the preset thresholds to recognize falls and some daily activities. All suspected falls were uploaded to the cloud server, which reduced wireless data transmitted. On the server, 13 kinds of features which can characterize different stages of fall were extracted. Then, the extracted features were standardized. Finally, SVM optimized by the GridSearchCV was used to make the terminal decision on the uploaded suspected falls. The experimental results showed that the specificity of the TBM stage reaches 65.23% when the sensitivity is 100%, which could significantly reduce the uploaded suspected fall events. The accuracy of the fall detection algorithm reaches 97.96%, the sensitivity and specificity reach 98.56% and 97.76%, respectively. Compared with the existing fall detection methods, the method designed in this paper has the advantages of low power consumption and high accuracy.