Abstract
This paper presents a novel machine-learning-based method for bed-leaving detection using Elman-type Feedback Counter Propagation Networks (EF-CPNs), which is particularly effective for processing time-series signals. In our earlier study, we have proposed a method based on CPNs, a form of supervised model of Self-Organizing Maps (SOMs), to produce category maps to learn relations among input and teaching signals. In this study, we introduce a feedback loop in CPNs as the second Grossberg layer so that the time-series features can be learnt. Moreover, we develop an original caster-stand sensor using piezoelectric films to measure, via bed legs, weight changes of a subject on a bed. The developed sensor has the features that it does not require a power supply for operations and can be easily installed on existing beds. We evaluate our sensor system by examining 10 people in an environment representing a clinical site. The mean recognition accuracy is 81.0%, while the mean recognition accuracy for the most important behavior terminal sitting is 98.0%. In view of the fact that most falsely recognized patterns belong to the categories of sleeping and sitting which are not so important for bed-leaving detection, we believe that the developed system can be applied to an actual environment as a novel sensor system requiring no restraint of patients.
