A watch-type human activity detector for the aged care

Human activity recognition is widely researched in the various filed these days. For the aged care, the one of the most important activities of old people is fall, since it causes often serious physical and psychological results. Many researchers have studied human activity recognition techniques in various domains; however none released to a commercial product satisfying the old people requirements, which are comfortable to wear it, weight-lighted and having exact accuracy to detect emergency activity and longer battery durance. Thus, to address them, we propose a systematic evaluating procedure for getting best minimum feature sets and best classification accuracy. We also do experiments for comparing the two features reduction techniques and four classification techniques in order to discriminate four each basic human activities, such as fall for the aged care, walking, hand related shocks and lastly general activity which is just last all arbitrary behaviour except existed three other patterns.