A novel video retrieval method to support a user’s recollection of past events aiming for wearable information playing

Abstract

Our system supports a user’s location-based recollection of past events with direct input such as in always ‘gazing’ video data, which allows the user to associate by simply looking at a viewpoint, and providing stable online and real-time video retrieval. We propose three functional methods: image retrieval with motion information, video scene segmentation, and real-time video retrieval. Our experimental results have shown that these functions are effective enough to perform wearable information playing.