摘要
为实现广播电视敏感人物的实时监测,将人脸识别、场景文字识别、语音识别等模型进行融合增强,设计建设了一套多模型融合增强的监测系统,对电视节目敏感人物进行自动监管。实践过程中发现,多模型融合增强能够显著提升敏感人物检测的准确率与召回率。实际应用效果表明,建设的广播电视敏感人物智能监测系统能够快速发现播出内容中的敏感人物,极大地提高了监测的速度和准确性,在很大程度上减轻监测人员的工作负担。
In order to realize the real-time monitoring of sensitive characters in broadcast and TV programs, this paper integrates and enhances the models of face recognition, scene text recognition, voice recognition and so on, and designs and constructs a set of multi-model fusion and enhanced monitoring system to automatically supervise sensitive characters in TV programs. In practice, it is found that multi-model fusion enhancement can significantly improve the accuracy rate and recall rate of sensitive character detection. The practical application results show that the intelligent monitoring system of radio and TV sensitive persons constructed in this paper can quickly find the sensitive persons in the broadcast content, greatly improving the speed and accuracy of monitoring, and greatly reducing the work burden of monitoring personnel.
作者
王戎
WANG Rong(Zhejiang Radio and Television Monitoring and Evaluation Center,Hangzhou 310012,China)
出处
《中国有线电视》
2021年第1期44-47,共4页
China Digital Cable TV
关键词
敏感人物
人脸识别
场景文字检测
语音识别
sensitive person
face recognition
scene text detection
speech recognition
