Human Fall Detection using Gaussian Mixture Model and Fall Motion Mixture Model

Deaths from accidental falls are the second leading cause of mortality from nonfatal injuries. Over the past two decades, there has been an explosion in research into the possibility of using technology to detect human falls on its own. This is an intriguing puzzle, and there are many different ways to solve it. Many various approaches have been presented during the past few years. These resources offer methods for identifying a wide variety of human motions, including walking, running, leaping, jogging, falling, and many others. Since falling is a common and potentially hazardous event for persons of all ages, but has a higher adverse effect on the senior population, prevention and detection of falls is particularly important among all these activities. Many existing fall detection systems cannot reliably reduce false positives due to the computational complexity of the algorithm they utilise. Instead of using wearable sensors, video cameras are the best option for fall detection. This article presents a visual method for recognising activities and detecting falls. The proposed method identifies falls only through the use of video-camera pictures, with no need for additional ambient sensors. Making a GMM (Gaussian Mixture Model) first. Using a fall motion vector, the URFall dataset's fall events are easily detected with minimal training time using a method called Fall Motion Mixture Model (FMMM), which is an implicit extension of GMM.