Research Of Wearable Fall Detection Perception Control System

As the proportion of elder people of our society grows bigger,children's dependency ratio grows higher,it means more and more elder people can not get comprehensive care,the phenomenon that elder live alone will be more common,and Fall is the leading killer,take great burden to their family.As a result,the research of accuraty perception of the older's fall has important significance.This paper mainly considers portability,comfort and sensitivity of fall detection device,designed a wearable perception system.In order to let the old man wearing more comfortable,using one perception device to collect human motion data,when system analysis judgment in a fall motion,perception device start the alarm system to inform nearby people to rescue the old man.Firstly,we investigate the existing fall detection algorithms and these implement methods,analyze their merits and faults.Designed a fall detection system based on one perception device.According to the system requirements,designed the hardware and software,using the subarea node perception device collect three axis acceleration and angular velocity of human motion,and preprocessing the data.Secondly,extracted the time sequence perception datato extracte the feature vectors,according to the classification performance of feature vectors,selected the better feature vectors to establish the vector space.Using different classification algorithms to classify the perception data of this paper,choice the best classification algorithm.Finally,analysis sensitivity of different kernel functions,used penalty factor to solve imbalance Sample problem,aimed at optimizating classifier,establish target function based on the classification forecast error rates,using Firfly Algorithm to optimize classification algorithm based on target function,improved the classification accuracy,and proved its validity by corresponding experiment.