Fall detection using visual cortex bio-inspired model for home-based physiotherapy system

Abstract

A home-based physiotherapy system aim to provide support for rehabilitation patients in home environment. Home-based physiotherapy system is useful for those who do not like to travel and lives in rural areas. The therapist can just monitor the physiotherapy exercise via online. This paper focuses on walking exercise monitoring and analyses the motion of the patient to detect any abnormality movement. The patient’s movement was captured by the low cost camera and this paper proposed visual cortex bio-inspired methodology to analyses the motion patterns. Visual cortex bio-inspired models mimics human vision system, mainly the primary visual cortex V1 and MT layer. The motions patterns are encoded using 3D Gabor spatio-temporal filter in V1 layer. Then generated spiking neuron model in MT layer formed active motion map based on Gaussian distribution. The extracted active motions plan was formed according to the direction, orientation and speed of the object movement. Then, the SVM classifier will identify the patient state either as a normal or as an anomaly movement. The robustness and accuracy of the system has been extensively tested with rigorous dataset that contained various types of fall. It can be scalable to become an online monitoring system that links with the rehabilitation centre for better support.