Fall Prediction and Detection through Inferencing at the Edge using the Pacific Research Platform
Kubernetes cluster

Among older adults, falls are the leading cause of injury-related morbidity and mortality. In 2015 29,000,000 falls were reported, resulting in 33,000 deaths. Imperial County is home to 22,000 adults age 65 and over; 80% are Hispanic. In 2013, falls were the second leading cause of unintentional injury deaths in Imperial County, and 657 people were hospitalized due to falls that year. Falls were the most common cause of non-fatal, unintentional injury hospitalizations, with the majority occurring in adults age 65 and over.

We are developing a wireless, wearable, low-power fall detection sensor to predict an imminent fall, and detect when a fall occurs, for elderly Latinos. The sensor will continuously acquire quantitative
metrics of physical activity (duration of walking, standing, sitting, and lying) and use this data to assess fall risk in Latino populations. While 28-35% of all community-dwelling people over 64 years of age fall each year, the frequency of falls among older Hispanic adults was found to be much higher. A 2013 study found 54% of participants in a Latino population at a community center had fallen in the last year, and 81% of those who fell were afraid of falling again. Fear of falling is a persistent worry that causes an individual to avoid physical activities and this fear contributes to the likelihood of further falls. Continued avoidance of physical activity contributes to a degradation in physical strength, agility, and balance, which increases risk of falling. Lack of physical activity among Latinos is correlated with, and may be a causal mechanism of, depression 40. Thus, the fear of falling in older Latino populations could lead to depression, which contributes to a health disparity between Latinos and the general population. Sensor wearers determined to be of high risk for falls can be referred to an evidence-based falls prevention program to acquire skills to reduce fear of falling, improve balance, and increase physical activity, thereby reducing the frequency of falling disparity between elderly Latinos and non-Latinos.

IEEE Globecom2019 Human Fall Dataset for paper Optimal Location for Fall Detection Edge Inferencing

[download] (253 MB)

Fall data captured using Noraxon IMU sensors which recorded 183 kinematic features that measure orientation of anatomical (joint) angles and limb linear acceleration with a 200 Hz sampling rate. File is a GNU Zipped tar archive of CSV format Noraxon files.