Classification Different Types of Fall For Reducing False Alarm Using Single Accelerometer

Abstract

Fall is one of the major causes of serious injury, which include fractures, traumatic brain injury, and death, to the elderly. True fall detection in time will improve the chances of survival and increases the likelihood of normal behavior recovery by up to 80%. Many researchers use accelerometers to detect fall as its convenience, low power and portable. However, simple threshold method can lead to false alarm in several ADLs (Activities of daily living) such as lying or sitting and several types of fall. This paper presents a fall detection algorithm to reduce false alarm using predetermined multi – thresholds in three phase of fall events. The performance of this techniques is evaluated using signals generated during in lab experiments that record the user’s movement signals during normal activities (walking, up/down stairs, standing up, sitting down and lying down) and a variety of fall cases. It was found that our method is able to classify 6 different type of fall and 6 ADLs with the accuracy is 92%, which was comparable to other methods.