Due to the rapid development of broadcasting and communication, large-capacity high-quality images such as 4K/8K UHD contents are provided in various forms. In particular, users' demand for ultra-low latency, uninterrupted and stable service provision for immersive media contents is increasing. However, since the transmission bandwidth and the number of channels are not sufficient to use only the terrestrial broadcasting network to provide large-capacity, high-quality content, there is a limit to satisfying users' requirements.
In order to overcome these limitations, a technology for converging and transmitting a terrestrial broadcasting network and a high-speed communication network such as 5G is required. In particular, it is necessary to transmit large-capacity, high-quality content based on IP. Among the 2nd generation digital broadcasting methods, ATSC 3.0 is a packet-based IP-based transmission technology that can provide broadcasting services that combine broadcasting networks and communication networks. In a situation where a broadcasting signal is unstable during transmission of large-capacity, high-quality content, service provision may not be smooth depending on the traffic of a communication network.
In order to solve this problem situation, in the actual field, IP network stream switching technology of UHD service uses a method of switching to a spare input when a failure of the main input occurs for two inputs. This method causes frequent switching, which not only shortens the lifespan of the equipment, but also causes disconnection, delay, and stoppage in the service provided.
In order to solve this problem, an AI-based multi-stream changer platform is required. When system devices fail due to transmission errors, it is possible to automatically judge and switch multi-input signals without dropping or changing signals by predicting or quickly determining the time of error occurrence for smooth stream switching.
Therefore, in this paper, I present an abnormal multi-stream detection technique based on data of various problematic situations in multi-stream changers used simultaneously in terrestrial broadcasting networks and high-speed wired/wireless networks and also present a platform that can verify it.
By analyzing packet loss before and after the reference point set in the multi-stream changer to which the proposed abnormal multi-stream detection technique is applied, it is expected to solve problems such as reliability of the equipment and broadcast accidents and can be predicted in advance to enable more stable and smarter switching than before.