LAUSR

Fall accidents are serious problems, especially for elderly people living alone. This is why many fall event detection methods have been proposed in the past. When constructing multiple classification models, a method is expected to be constructed for every fall and nonfall actions. However, in the context of IoT, it would be challenging, if not impossible, to cover all scenarios, because it is necessary to keep the computational cost low. Thus, we propose a hidden Markov model that aggregates the actions measured with a single Doppler sensor (based on the likelihood) in each fall and nonfall events. To validate the proposed method, we conducted an experiment with 20 subjects who performed three fall actions and four nonfall actions each. An accuracy of 0.95 was achieved, and the number of generated fall and nonfall detection models was 1:2 and 1:3, respectively. The experiment’s results show that the proposed method has successfully reduced the number of classification models, which is crucial for IoT applications.