Review of fall detection techniques: A data availability perspective

Abstract

A fall is an abnormal activity that occurs rarely; however, missing to identify falls can have serious health and safety implications on an individual. Due to the rarity of occurrence of falls, there may be insufficient or no training data available for them. Therefore, standard supervised machine learning methods may not be directly applied to handle this problem. In this paper, we present a taxonomy for the study of fall detection from the perspective of availability of fall data. The proposed taxonomy is independent of the type of sensors used and specific feature extraction/selection methods. The taxonomy identifies different categories of classification methods for the study of fall detection based on the availability of their data during training the classifiers. Then, we present a comprehensive literature review within those categories and identify the approach of treating a fall as an abnormal activity to be a plausible research direction. We conclude our paper by discussing several open research problems in the field and pointers for future research.

Remote patient monitoring: a comprehensive study

“…Another very recent review on fall detection techniques further narrows down the conversation to 'a data availability perspective' (Khan and Hoey 2017). They argue that sufficient data on falls may not be available during training classifiers of the systems that use supervised machine learning methods.…”

Ambient Technology to Assist Elderly People in Indoor Risks

“…Falls are the major cause of fatal and non-fatal injuries among elderly people, which creates an obstacle for independent living [9]. According to the World Health Organization (WHO) [10], the frequency of falls increases with age and frailty.…”

“…We illustrate the framework of assistive technology for fall risk in Figure 4. Khan et al [9] present a taxonomy of fall detection techniques based on two high-level categories from the data availability perspective: (I) sufficient training data for falls; and (II) insufficient or no training data for falls. The first category (I) of this taxonomy presents the case where sufficient data for falls is available to train the classifiers.…”

Three Ways to Improve the Performance of Real-Life Camera-Based Fall Detection Systems

“…An automatic fall detection system which does not need user intervention can overcome this problem. Because of the importance of this issue, a lot of research has been done to solve the fall detection challenge, as can be seen in the numerous available review articles [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19]. There are different ways to detect a fall; Mubashir et al categorizes them, for example, in three categories: wearable sensors, vision, and ambient/fusion [14].…”

“…These conditions are underrepresented in most simulation data sets (DS). As almost all published research is only validated using simulated data [16,28], often these challenges remain untested. In our previous work [28], we showed that our fall detection algorithm that was based on a simple background subtraction method performed similarly as the state-of-the-art on a publicly available simulation data set.…”