Human Activity Sensing and Applications to Medical and Caregiving Domains
Int'l Sympo. Learning Analytics And Adaptive Aids for Active Adults At All Ages (LA9)
In this talk, we introduce activity recognition technology using mobile type sensors including smartphones and applications to medical and nursing fields. Although a lot of activity recognition technology has already been introduced, there are many challenges in collecting realistic datasets and developing algorithms for complicated and long-term activities. To this problem, from our research, we proposed a method [UbiComp2015] that uses prior knowledge of the activity segment of the day, and a method that automatically corrects when the timing of the labels are inaccurate [MobiQuitou2016]. We also proposed a method [MobiQuitous2016] that corrects differences among individuals by transfer learning. Along with these works, we talk about the future prediction of nursing work volume and patient prognosis in combination with medical data in the hospital[UBI16], and talk about the trial of recognizing whole staffs' activities in nursing homes for 4 months [UBI17].