Streaming media processing method and device based on online reinforcement learning and electronic equipment

Abstract

The embodiment of the invention provides a streaming media processing method and device based on online reinforcement learning and electronic equipment, and the method comprises the steps: obtaining streaming media state data of a current time period; through an online machine learning model, according to the streaming media state data, determining streaming media control information of a streaming media client in the current time period and sending the streaming media control information; and obtaining streaming media state data of a next time period, training the online machine learning model, and updating model parameters of the online machine learning model. According to the embodiment of the invention, the online machine learning model is used, the streaming media control information is determined according to the current streaming media state data and is transmitted to the client to control further streaming media processing operation, and iteration operation is carried out in this way, so that the method adapts to a dynamically changing network environment. Meanwhile, by training the online machine learning model, model parameters are continuously updated so as to continuously adapt to a new network environment.

Description

technical field [0001] The application relates to a streaming media processing method, device and electronic equipment based on online reinforcement learning, which belongs to the field of computer technology. Background technique [0002] In order to maintain a high-quality user experience (QoE, Quality of Experience), traditional interactive streaming media applications use rule-based protocols, such as congestion control at the transport layer and video bit rate adaptive algorithms at the application layer . However, rule-based protocols cannot adapt to highly heterogeneous modern Internet networks, including cellular / WiFi wireless links, long-distance fiber optic links, cloud-based data center links, etc., all of which have different bandwidth, latency, and buffering capabilities . [0003] In recent years, data-driven machine learning algorithms have been proposed to improve the QoE of streaming media. However, existing solutions usually adopt the "learn offline, run...

AI Technical Summary

Problems solved by technology

However, since it is difficult for simulators or emulators to faithfully simulate the dynamics of streaming interactions in the complex real network world, such offline learning models perform poorly when applied to the real world.

Images