Research Of Wrist-worn Fall Detection System Based On Sensor Data Fusion Technology

With the age grows, the old man whose body function is declined and response is slow will face more and more risk. The rate of injury or even death by falling down in the elderly is increasing dramatically. Therefore, the lives of elderly people need to be concerned and cared. However, the elderly people always live alone with no offspring around. In addition, elders who may go out to walk around, buy foods, do exercises and other activities are hard to monitor. Therefore, it is of great significance to develop a device which can monitor the physical condition and location of the elderly at any time and at any place. Such a device must be able to give an alarm and find the old man timely when they are falling or lost.As the old people are not familiar with the mobile phones and other high-tech products,a wrist fall detection system is designed in this paper according to the characteristics of the daily activities of the elderly. Besides, by analyzing the current fall detection technology, a secondary judge fall detection fusion algorithm is proposed to apply to the fall detection system. This paper mainly completed the following works:Firstly, we build a fall detection system hardware platform. The system is a independent device which is worn on the wrist of the human body to detect the position of falling action. This device is also suitable for the chest, waist and other locations, so that it is easy to wear for the elderly. The system uses an advanced six-axis inertial sensor. The acceleration sensor and the gyro sensor are integrated in a single chip. It can greatly save the board level space and the cost. The hardware uses a low power MCU and an optimized battery power management program to achieve portability, miniaturization and low power consumption.Then, according to the characteristics of the system, we fuse the sensor data to select multiple features and achieve dimensional feature vectors extracted by PCA. In order to better distinguish between falling and daily activities, this paper presents a secondary judge fall detection fusion algorithm. This algorithm is based on the threshold value orientation and pattern recognition classifier. It can improve the accuracy and stability of the system.At the same time, we implement a prototype of the fall detection system and test the detection rate of algorithm.Experimental results show that the algorithm reduces the rate of false-alarm and negligent-alarm and improves the detection performance of the system.