Article Preview
Top1. Introduction
The population of aging people is increasing year after year. It raises the concern for providing extra health care. Petry & Yager (2012) illustrated the need of human-agent communication with robotic assistance for elderly care. The health condition of the aged adults gets aggravated with the shift in the locality of their living. For their well-being, the concept of aging in place is encouraged to empower the aged adults to live independently in their own place (Belshaw et al., 2011). To improve the quality at their place, it is necessary to monitor their health and safety conditions. Fall of the person is of major concern in the elderly care. Elders suffer fall due to their reduced ability in taking personal care and protection.
Marquis-Faulkes et al., (2005) stated that around 60,000 human falls are reported every year. Their paper gives the study of an active ambulatory institutionalized population of adults over the age of 65 years. It revealed that an annual fall rate of 668 incidents per 1000, with an increase in frequency for successive age groups above the age of 75 years. For elders above 79 years, fall is the second leading cause of unintentional-injury death. Forty five per cent of all people under study suffered at least one fall during the study period. Accidental fall results in physical injuries that lead to major health complications both physically and mentally. Fear, anxiety and depression also increase due to fall. As cited by Perolle et al. (2006), 30% of the older persons fall at least once a year and 70% of accidental deaths in aged persons are due to fall. Hence it is necessary to have reliable automated fall detection system to detect all events. There are two existing detection systems. First kind of fall detection systems, identify/classify the fall in an autonomous way. The second system expects a notification from the fallen person. The later type of the fall detection system has the disadvantage when the fallen person is unable to give notification due to loss of consciousness.
The prevailing method for fall detection uses accelerometer or gyroscopes (Hemalatha & Vaidehi 2013) to identify day to day activities like sitting, standing, walking etc. The results obtained by these sensors are more erroneous as some activities result in high acceleration which leads to false alarms. Hence vision-based fall detection system is proposed to give more accurate information about the person activities in indoor environment. Fall detection using camera sensor gives accurate and reliable results. Images from camera can be used to generate timely alert and provide complete information about the position and nature of fall. It aids in providing immediate health care and service to the needy. A vision-based fall detection system is shown in Figure 1. Computer vision system provides better solution for fall detection. The person is continuously monitored through image network and the images are analysed from remote place.
Figure 1. Vision based fall detection system
The challenges associated with fall detection are as follows:
- •
In an indoor environment, the Field-of-View (FoV) is obstructed due to obstacles such as furniture, household appliances etc., which leads to occlusion.
- •
In home environment, living places are not equally illuminated.
- •
Self-occlusion condition.
- •
Availability of the training samples for fallen person is difficult as the person may fall in different postures.