MEF融合HFF的戏剧视频关键情节自动提取

为了更好地从戏剧视频提取关键情节,提出了一种基于音乐情感特征(MEF)融合人脸特征(HFF)的自动提取方法.首先,利用基于音频指纹技术的二级音乐情感识别方法进行音频识别,并利用人脸特征进行视频识别;然后,利用音频和视频识别得到的各元素获取关键情节值,从而提取关键情节;最后,提出了一种量化评估方法评估关键情节提取方法的一致性.在四个戏剧视频上的评估实验验证了该方法的有效性及可靠性,相比其他几种较新的提取模型,该方法提取效果更好.

Automatic highlights extraction for key plot of drama video based on fusion of MEF and HFF

In order to extract key plot of the drama videos, an automatic extraction for key plot based on fusion of music emotion features (MEF) and human face features (HFF) is proposed. First, we use a two music-based emotion recognition method based on audio fingerprinting technology to do audio recognition, and use human face features to do video recognition. Then, use the elements of audio and video identification to get access to key plot value to extract key plot. Finally, a quantitative assessment method is proposed to estimate the consistency of proposed method. The quantitative evaluation on the four drama videos method gives the validity and reliability of the assessment, while compared with several other new extraction methods, the proposed method achieves better results.