Retrieval of personal experience records using relevance feedback with a structuring filter

In this paper, we propose a novel method of "relevance feedback with a structuring filter" that efficiently retrieves personal experience records in the form of videos taken by a head-mount camera or other devices. Because such data can be redundant and shaky, they are not suitable for thorough perusal and we therefore require an efficient method of browsing and retrieving these data. Relevance feedback is a technique that supports flexible content retrieval; however, it is often difficult to select from a vast amount of data appropriate samples that are similar to the records being sought. To solve this problem, our method associates contents based on semantic structures, and returns closely related portions in conjunction with portions obtained by relevance feedback. This process makes content retrieval more flexible and efficient. In this paper, we define the semantic structures of personal experience records, describe the relevance feedback technique with a structuring filter, and present experimental results that show the efficacy of our proposed framework.