| China will become the countries with high levels of population aging by2030, andit showed a trend of accelerated development. Fall not only damage the old body andthreatening the safety of life of the elderly, but also resulting in a decline to the elderlyliving independence. Therefore, automatically detect fall behavior and promptly reportto the others for help will greatly reducing the damage to the body and improving thequality of life of the elderly.Because off all behavior causes acceleration and human inclination’s mutations,this paper presents a fall detection technology based on acceleration and humaninclination inclination’s change and designs a wearable fall detector. A support vectormachine (SVM)-based pattern recognition fall detection algorithm has been realized inthis fall detector. Human inclination data was got by calculating the acceleration.Extracting and calibrating the feature from the acceleration and the human bodyinclination data. In order to gets support vector machine (SVM) parameters, thenormalized eigenvectors be input to the classifier training modules and be trained.Through data filtering, feature extraction, calibration, and other operations, inspectedsamples were input to classifier model for testing; if it is positive, gives alarm.Wearable fall detector is mainly composed of two parts of embedded hardware andapplication software. In STM32embedded hardware systems, uC/OS-II embeddedsystem software was using to design the fall detector based on support vector machine.At the same time, participants with different fall actions are testing theperformance of the wearable fall detector. This paper analyzes the interval distributionof features of the test staff’s front, rear, left, right four directions falls behavior andnormal behavior and analysis the features similarity of fall behavior and violentmovement. A large number of experimental data show that the proposed fall detectionalgorithm have high detection accuracy to fall behavior and normal behavior. |
PDF Full Text Request