Article Preview
TopIntroduction
In recent years, the trend of living alone has been increased which exposed elderly persons of families to be unaccompanied. This phenomenon increased the risk of falling especially in elderly people or people suffering from diseases like Alzheimer's, dementia, diabetes, etc. Among all these Alzheimer is a major cause of falls in elderly people. According to Global QYResearch via COMTEX report “the global fall detection systems market was valued at USD 365million in 2018 and is expected to reach a market valuation of approximately USD 544 billion by 2026 growing at a CAGR of 4.2% during the forecast period” (Global QYResearch 2019). Human fall can be stated as suddenly going down towards the ground unintentionally due to any disease, unconsciousness, and weakness, etc. or activity as a result of an accident. Falls can cause severe physical and psychological injuries to the person including death. In addition to the physical disturbance people also suffer from fear, anxiety, physical injuries, expensive medication, long hospital stay, social burden, etc. If immediate help does not reach the older people then it can result in fracture and long lie and further delay may even cause the casualty. Human falls detection is becoming a big challenge to encounter effectively.
Sensors are an essential component for the detection of human falls. Many different types of sensors are available like accelerometer, gyroscope, camera, vibration, etc. These sensors can be categorized into three different types (1) wearable sensors (2) ambient sensor and (3) vision-based sensors. The wearable sensor is those sensors that can be worn around the waist, neck, and wrist. Some authors have proposed wearable based fall detection methods(Sztyler, Stuckenschmidt, and Petrich 2017; Zhang, Wang, et al. 2006). Ambient sensors imply vibration sensors, acoustics sensor or a combination of them(Noor, Salcic, and Wang 2017). Vision-based sensors are considered most non-obtrusive sensors but they are not easily portable whereas the wearable sensor can be easily worn and easily portable. Since the last few decades, the induction of smartphones among the daily living of people and advanced technologies have created a favorable environment for the growth of the fall detection system. The working of fall detection systems can be understood as shown in Figure 1. IoT technology provides a simple platform to capture human fall detection efficiently. As the algorithms are very complex to run in smartphones, therefore, data is collected via smartphone and processed at cloud environment.
Falls can be detected using two ways: using threshold values(Abbate et al. 2012; Wang et al. 2014) and by using machine learning techniques(Albert et al. 2012; Zhang, Wang, et al. 2006). In threshold-based approaches, calculated values of resultant are checked, if it is between lower and upper threshold values then it is daily living activity otherwise it is fall. Although machine learning algorithms are more complex than threshold-based methods but they provide better results than threshold-based methods. Machine learning-based fall detection algorithm uses various techniques like neural networks (Musci et al. 2018), classification algorithms, deep learning (Shin et al. 2016)(Ma et al. 2014), etc.