Research And Implement Of Fall Detection Algorithm Based On Android Platform

As the global population continues to increase the degree of aging,human fall detection,early warning and protection technologies,as the elderly to prevent accidental fall injury and improve the quality of life an important guarantee.Wearable computing and human sensor network technology continues to mature,through scientific means to identify the elderly fell to become a new hot spot at home and abroad.This technology not only in theoretical research has a high significance,but also in the practical application of the value is very large.The research of fall recognition problem is divided into four parts,the establishment of fall model,the design of experimental scheme,the processing of raw data and the application of discriminant algorithm.The acceleration and angular velocity in the process of human motion are selected as the main feature quantity,and the algorithm of threshold value,support vector machine and decision tree are used to realize the fall detection or early warning.Fall detection is essentially a simple binary classification problem,our main goal is to distinguish between an activity falls or daily behavior.Threshold method is relatively intuitive,the algorithm is relatively easy to implement,is the most common research methods.Because there is no definite model at present,the size of the threshold is mostly based on empirical judgment or statistical experimental results,its deficiency is that the setting of threshold has great influence on the recognition result,and the stability is poor.The method of pattern recognition has great advantages for large sample classification.However,in practical research,it is difficult to obtain a large number of true fall data,mostly based on small sample.Support vector machine has certain advantages to solve the problem of small sample and non-linear pattern recognition.In this thesis,we focus on the motion feature weighting method based on support vector machine to achieve fall detection.Firstly,the paper analyzes the output data of the smart phone sensor,selects and extracts the feature quantity which differentiates falls and daily activities,constructs the virtual samples of falls and daily behavior activities by using Gaussian distribution principle.Then,a fall recognition model based on constructed sample set is established to verify the generalization performance of the algorithm.And the algorithm constructed in this paper is implemented on the mobile phone platform of Andrews,and the practical application effect of the fall detection algorithm is verified.The experimental results show that the proposed method can detect 90.67% sensitivity,92.96% specificity and 92.14% accuracy,and the performance of the proposed algorithm is verified.But also the side validation of the platform based on Andrews fall detection software robustness and stability.