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

Abstract

More than thirty percent of persons over 65 years fall at least once a year and are often not able to get up again. Camera-based fall detection systems can help by triggering an alarm when falls occur. Previously we showed that real-life data poses significant challenges, resulting in high false alarm rates. Here, we show three ways to tackle this. First, using a particle filter combined with a person detector increases the robustness of our foreground segmentation, reducing the number of false alarms by 50%. Second, selecting only nonoccluded falls for training further decreases the false alarm rate on average from 31.4 to 26 falls per day. But, most importantly, this improvement is also shown by the doubling of the AUC of the precision-recall curve compared to using all falls. Third, personalizing the detector by adding several days containing only normal activities, no fall incidents, of the monitored person to the training data further increases the robustness of our fall detection system. In one case, this reduced the number of false alarms by a factor of 7 while in another one the sensitivity increased by 17% for an increase of the false alarms of 11%.