Action recognition is the technology of understanding people’s behavior and classification from video or image sequences. This thesis uses the deep learning approach for action recognition to realize webcast supervision. This paper uses the convolutional neural network (CNN) and the Gaussian Mixture Model (GMM) to establish the webcast supervision system. At the same time, streaming-based wireless communication network technology is adopted to ensure video transmission speed and quality. Results show that the average detection speed of the system can reach 11.86 frame/s, and the average recognition accuracy is 92.16%, and the missed detection rate is lower than 5%. The design of this system can fully meet the requirements of webcast supervision.

中文翻译：

---

基于卷积神经网络和无线通信的网络广播监管研究

动作识别是从视频或图像序列中理解人的行为和分类的技术。本论文采用深度学习方法进行动作识别，实现网络直播监督。本文使用卷积神经网络（CNN）和高斯混合模型（GMM）建立网络直播监督系统。同时，采用基于流媒体的无线通信网络技术，保证视频传输速度和质量。结果表明，系统平均检测速度可达11.86帧/秒，平均识别准确率为92.16%，漏检率低于5%。本系统的设计完全可以满足网络直播监管的要求。