Sub-Project B: Wearable Fall Event Sensors
Research team: Ed Park (co-leader), Stephen Robinovitch (co-leader), Greg Mori
In this Sub-project, we are combining miniature sensors (Microstrain G-Link and APDM Emerald sensors) with custom-designed data analysis algorithms to create wearable fall recorders that can accurately detect and log the onset of a fall, and record key characteristics associated with the initiation, descent, and impact stages of falls. The first stage of this work involves laboratory experiments with young adults, who act out a variety of daily activities, as well as falls and near-falls. By using a large array of sensors in these experiments, we can test how features of the sensor array (number of sensors, mounting locations, and type of sensor signal) and the type of data analysis algorithm affect the sensitivity and specificity of the system for detecting key characteristics of falls and near-falls. Simultaneously, we will conduct focus groups and interviews with LTC residents and care providers to identify the factors that influence individuals’ willingness of to wear these devices, and care providers’ attitudes in supporting their use. This will allow us to select the hardware and data analysis algorithms for the next stage of the research, which involves using the system to measure falls, near-falls and mobility patterns in LTC residents.
Figure 3. TIPS trainees Omar Aziz (right) and Thiago Sarraf conduct experiments in the IPML laboratory to test the ability of wearable sensors to distringuish the cause of falls.
Recently, we conducted falling experiments in the IPML laboratory to determine how the number and location of 3D accelerometers affects the accuracy of a machine learning algorithm to distinguish the cause of falls. 16 participants underwent falls and near-falls due to slips, trips, simulated faints, and loss-of-balance due to weight shifting. We found that data from a single accelerometer (placed at the waist) provided a minimum sensitivity of 76% to distinguish fall type; accelerometers at the waist and sternum increased this to 93%, and accelerometers at each ankle, and a third at the waist or head increased this to 97%.