Discriminative Data Visualization for Daily Behavior Modeling
Masamichi Shimosaka, Taketoshi Mori, Akinori Fujii, Tomomasa Sato · Advanced Robotics · 2009
This paper describes a novel data visualization technique for intuitive interpretation of human behavior modeling results. Our visualization method captures the relational structure of clusters obtained by some human behavior modeling method, and reflects it to two- to three-dimensional space. The main advantage of our method is that the method can be applied to massive daily sensor log analysis. This paper reports some experimental results on pyroelectric sensor log analysis where the data spans months to years.