Improving the performance of camera-based fall detection systems

More than thirty percent of persons over 65 years fall at least once a year and are often not able to get up again, making this a major hazard for those living at home. Camera-based fall detection systems can help by triggering an alarm when falls occur. A new paper presents ways to reduce rates of false alarms in these systems.

Debard, et al. (2017) demonstrate three methods for reducing false alarm rates in camera-based fall detection systems:

  1. Using a particle filter combined with a person detector increases the robustness of foreground segmentation, resulting in a 50% reduction in false alarms.
  2. Selecting only non-occluded falls for training further decreases the false alarm rate on average from 31.4 to 26 falls per day. Most importantly, this improvement is also shown by the doubling of the AUC of the precision-recall curve compared to using all falls.
  3. Personalising the detector by adding several days containing only normal activities (no fall incidents) of the monitored person to the training data further increases robustness. In one case, this method reduced the number of false alarms by a factor of 7.

Other GRaCE-AGE research in fall detection systems includes the development of radar-based fall sensors. The GRaCE-AGE system, when integrated with an older adult's care network, is able to alert carers to a fall based on the output of connected fall sensors. This functionality can potentially improve health outcomes and save lives, while at the same time reducing the burden on the care network, but is dependent on accurate sensor hardware and analysis systems.

This research in reducing the rates of false alarms in these systems is an essential step in producing fall detection systems that are ready for market and able to make older people living at home safer, while reducing the burden on care services.

Glen Debard, Marc Mertens, Toon Goedemé, Tinne Tuytelaars, and Bart Vanrumste, “Three Ways to Improve the Performance of Real-Life Camera-Based Fall Detection Systems,” Journal of Sensors, vol. 2017, Article ID 8241910, 15 pages, 2017. doi:10.1155/2017/8241910

Available online: https://www.hindawi.com/journals/js/2017/8241910/