基于用户兴趣的视频片段提取方法

Video highlight extraction based on users' interests

视频精彩片段提取是视频摘要、检索等领域的热点研究问题。若自动提取视频片段能考虑用户个性化需求,将能在电影推荐、新闻摘要等实际应用场景中发挥巨大作用。针对基于用户兴趣的视频片段提取问题,提出了使用互联网图像的用户感兴趣视频精彩片段提取框架,使用近义语义联合组权重模型权衡由不用的描述词检索得到的存在差异但语义相近的图像组,根据图像组与视频内容的相关性为不同的图像组分配不同的权重。实验结果表明了所提出的方法对提取用户感兴趣内容的有效性。

Extracting highlights is of great interest in video summary,retrieval and other fields.If the automatic video clip can take into account the user's personalized needs,it will be able to play a great role in the actual application scenes such as movie recommendation,news summary and so on.For the extraction of user interest based video clip,a novel joint group weighting learning framework is proposed to leverage different but related groups of knowledges learnt from the Web images to videos.Under this framework,the weights of different groups are learnt in as a joint optimization problem,and each weight represents how contributive the corresponding image group is to the knowledge transferred to the video.The experimental results show the effectiveness of the proposed method to extract the content of interest of the user.