| 主权项 |
1. A method for a personalized video recommendation, comprising:
detecting, by one or more hardware processors, a viewing activity of at least one user of a content-presentation device capable of presenting multiple video programs in one or more channels; representing, by the one or more hardware processors, user interests of the at least one user by using a topic model, including:
performing visual and textual analysis to each video to generate a codebook for each video, andmodeling a topic distribution of each video and a word distribution of each topic; discovering, by the one or more hardware processors, the user interests from user viewing histories; generating, by the one or more hardware processors, a personalized video list of personalized video contents based on the user interests; recommending, by the one or more hardware processors, the personalized video contents to the at least one user; and delivering, by the one or more hardware processors, the recommended personalized video contents to the at least one user such that the personalized video contents are presented on the content-presentation device; wherein discovering the user interests further includes discovering the user interests at different granularity as a mixture of topics by using a Markov chain to capture a time-varying property of interests and the topic model to calculate the topic distribution of each video using both textual and visual information; and wherein performing the visual analysis including:
dividing each video into a plurality of shots,determining at least one keyframe of each shot, wherein each keyframe is represented by a plurality of local patches,detecting visual features from the keyframes of the plurality of shots, andgenerating a visual codebook for each video based on the visual features. |
