Novel sensing technology and system help predict falls and prevent injuries to seniors

KAIST researchers are developing novel motion sensing technologies and wearable systems to predict falls and prevent fall-related injuries for seniors.

Article | Fall 2021

Falls in older people are a major public health concern due to their high prevalence, serious consequences, and heavy burden on society. Around one in three older people falls each year, and 20~30% of falls result in injuries, as the leading cause of fatal and non-fatal injuries in older people. Due to the complex multifactorial nature of falls and the short period of falling, it is very challenging to predict a fall before it occurs and then to provide protection for the person who is falling.

In order to tackle this challenge and support safe aging, Prof. Shuping Xiong and his team in the Department of Industrial and Systems Engineering at KAIST have been developing motion sensing technologies and wearable systems to predict the falls and prevent fall-related injuries for the older people. Based on the motion data continuously recorded from a wearable inertial sensor on the waist, they have developed a novel hybrid deep learning model (ConvLSTM, Figure 1), which can accurately predict a fall during its initiation and descending but before the body impacts the ground (pre-impact fall) so that protective devices can be triggered in time to prevent fall-related injuries. The performance of the hybrid model was evaluated on a large public dataset of various falls and activities of daily living (ADLs), showing both high sensitivity (>93%) and specificity (>94%) for all three fall stages [1].

In addition, Prof. Xiong’s team constructed a new, large-scale open motion dataset KFall with accurate fall time labels (Figure 2), and proposed three different types of benchmark algorithms with different complexity for predicting falls before the body-ground impact [2]. The benchmark algorithms were comprehensively evaluated in terms of both accuracy and practicality measures (lead time, latency). This open large-scale motion dataset and benchmark algorithms could provide researchers and practitioners with valuable data and reference to develop new technologies and strategies on fall prediction and proactive injury prevention for older people.

The research team is currently developing a novel wearable on-demand smart airbag system to seamlessly integrate the developed pre-impact fall prediction algorithms with airbag inflation technology, which automatically deploys before ground impact for preventing injuries in the event of a fall. This is an ongoing research collaboration with research teams from Yonsei University and Safeware Inc., funded by Institute of Information & communications Technology Planning & Evaluation (IITP).

References
[1] Yu X, Qiu H and Xiong S*, 2020. A novel hybrid deep neural network to predict pre-impact fall for older people based on wearable inertial sensors. Frontiers in Bioengineering and Biotechnology, 8: 63.
[2] Yu X, Jang J, and Xiong S*, 2021. A large-scale open motion dataset (KFall) and benchmark algorithms for detecting pre-impact fall of the elderly using wearable inertial sensors. Frontiers in Aging Neuroscience, 13: 692865.