There is a growing global trend to prevent the acquisition of identifiable personal information. We aim to realize accurate action recognition and data tracking for the same person, even from anonymous sensing data, without facial images or personal matching. By factorizing sensing data into the features of humans, actions, and objects, we can eliminate the influence of each other on the recognition. This work contributes to the construction of anonymous big data that enables advanced data gathering, distribution, storage, and analysis.
- [Related project] Lesson Design (2011~)