Intelligent Monitoring System of the Elderly's Fall Based on Video Sequences
Ehsan Azizi, Ahmad Delbari · 2013
Every year thousands of the elderly suffer serious damages such as articular fractures, broken bones and even death due to their fall. In this paper, based on the analysis of images taken from the elderly's movement, an efficient system has been proposed that, in the first phase, simulates the movement of the elderly by detecting their abnormal walking. Thus by combining several important features, including an estimate of body angle, representation of the Motion History Image (MHI) and estimate of the magnitude and direction of movement, the speed of the falling is calculated. This system has been implemented on a set of 57425 video frames received from the elderly residing in Farzanegan Health Care Center in Mashhad city in Iran and the video sequences containing the actual occurrence the of falling. There were only 48 falls and 163 like falls on video sequences with mean age of 66.5±6.1 years. Out of a total of about 58 people including 43 men and 15 women, 39.65% were faced with falls at ambient of health center. All the sequences were randomly converted into four Movie categories with these details: AVI format, 120×160 pixels resolution and 15 fps. Compared to such techniques as Shieh and Rougier, 94.1% average accuracy (AAC), 92% sensitivity and 94.47% specificity indicate the ability of the system in identifying the incidents similar to the fall. The high speed data processing of the algorithm, 92.91% detection rate (DR) and insignificant false alarm rate (FAR), 5%, distinguishes the proposed system from similar ones, particularly due to its intelligent monitoring and its real time detection of the elderly's fall.