User behavior identification method and system for intelligent video conference

• Main Authors: MENG XIABING, LIU CHEN
• Format: Patent
• Language: Chinese; English
• Published: 22.08.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a user behavior identification method and system for an intelligent video conference, and relates to the technical field of artificial intelligence, and the method comprises the steps: constructing a space-time dual-support backbone network; constructing a space-time double-branch network; the space-time double-branch network comprises a target detection network YOLOv7, a face branch, an action branch, a whole-course time sequence branch, a splicing layer concat and two full connection layers; training the space-time double-branch network; and obtaining a to-be-identified conference video, and identifying the current action category of each participant in the to-be-identified conference video by using the trained space-time double-branch network. According to the invention, the accuracy and efficiency of conference participant behavior identification can be improved. 本发明公开一种面向智能视频会议的用户行为识别方法及系统，涉及人工智能技术领域，该方法包括构建时空双支骨干网络；构建时空双支网络；时空双支网络包括：目标检测网络YOLOv7、面部分支、动作分支、全程时序分支、拼接层concat以及两个全连接层