Fall Detection Technology And Application Based On Fusion Data Of Visual Image And Wearable Computing

The aging of the population is a common problem that all countries in the world are facing.With the continuous aggravation of the aging degree in China,the medical resource investment for the elderly population will continue to increase.At the same time,falls have been recognized as the most common dangerous behavior in the daily lives of the elderly.Therefore,the research of the fall detection system has significant significance in reducing the physical and psychological damage and medical costs of the elderly.At present,there are still many shortcomings in the solution to the fall detection of the elderly.The fall detection technology based on computer vision is more impressive in undisturbed scenes,but it is susceptible to environmental influences,and the use scene is limited,such as: being affected by background light,people are blocked by objects,etc.In addition,the fall detection technology based on wearable computing has the disadvantages of poor algorithm stability,low recognition accuracy,and difficulty in ensuring the sensitivity and specificity of the system at the same time.Aiming at the above problems,this paper proposes a new method,which combines computer vision and wearable computing data with fall detection algorithm.Firstly,the activity awareness module integrating the three-axis accelerometer,three-axis gyroscope and Bluetooth was designed and developed to collect the activity data of the elderly and transmit the data in real time.Secondly,the deep learning algorithm was used to extract the human body posture data from the image data collected by the camera.The collected human activity data and human body posture data are normalized and timed,and two network structures are designed to extract the features of the processed data,and the two features fusion.The activity recognition and fall detection of the fusion data were performed by constructing a neural network.Finally,the experimental platform is built and the algorithm is tested.The accuracy of the offline test of the fall algorithm is 99.2%,the average sensitivity is 99.5%,the average specificity is 99.8%,the online test accuracy is 98.9%,and the average sensitivity is 99.2%,the average specificity was 99.5%.Thus,it is proved that the fall detection using the fusion data of computer vision and wearable data has high accuracy and robustness.This topic provides technical support to help the elderly get timely assistance,reduce the damage caused by falls,and provides new ideas and guidance for the deep application of computer vision and wearable computing technology in daily life.