Information collection system for fall detection of stroke patients under cascade algorithm in the context of multi‐modal information fusion and e‐health

Abstract

AbstractTo effectively detect stroke patients after falling under e‐health, the fall situation of patients is graded, and warnings are given to improve the survival probability of stroke patients after falling. First, the fall model of stroke patients is analysed. According to the model, multi‐modal information fusion fall detection technology is proposed, including data fusion algorithm and feature recognition technology. Also, various sensor data are adjusted. The fall detection cascade algorithm is proposed to classify data with different features in order, thereby completing target detection. Finally, combined with heart rate sensor, height sensor and microcontroller unit software and patients' e‐health information, the research and development of an information collection system for fall detection are realized. Six young volunteers are selected to test the system performance. The results show that for the six testers, the heart rate detected by the ordinary device and the device of this investigation is the same when it is resting in different states of resting, walking, as well as walking and falling. The heart rate difference between walking and falling detection is not large (within the allowable error range of the device). But the best detection effect is to measure after the patient falls, which not only reduces power consumption but also keeps the detection error to a minimum. The height sensor is in the static state, increased by 75 m in the vertical direction and decreased by 75 cm from the static position in the vertical direction. The height difference of the data information exported from these three cases has some errors compared with the actual 75 cm. The tester's three situations of resting, walking and falling, standing up after sitting still and falling are observed. The waveform when resting is stable, and the acceleration information also fluctuates significantly when walking and after standing up. The accuracy of the developed system is above 80%, which achieves the expected effect and assisted medical treatment of fall detection based on patients' e‐health information. The results show that the proposed information collection system for fall detection of stroke patients under cascade algorithm in the context of multi‐modal information fusion has good performance and high practical value.