Abstract:In recent years, with the continuous development of network science and technology and the continuous promotion of “three networks convergence”, IPTV (Interactive Network Television) has shown a rapid development trend. IPTV is different from the traditional one-way broadcasting mode of television, it can achieve interaction with the audience, and provide more personalized and diversified videos. With the rapid development of new media, massive video resources and a large number of video-related information are sweeping in. The evaluation of video can no longer be limited to the ratings of traditional platforms. Video playback on new media platforms, network impact, video content and other related data are also important indicators affecting video evaluation. Video quality is closely related to video content characteristics. Different videos have different sensitivity to the same packet loss rate. The IPTV video quality evaluation model based on content features considers the video content characteristics. Firstly, the QP, bit rate and motion vector (MV) information of the video is obtained by analyzing the video stream. Then, the time complexity of the video is calculated by using the obtained MV, and the spatial complexity of the video is calculated by quantization parameters and bit rate. Videos are classified by clustering analysis. On this basis, the BP neural network is used to establish the model and evaluate the video quality, so as to better reflect the visual perception of the human eye. The model has low computational complexity and is suitable for IPTV video quality assessment with certain computing power in network nodes.

IPTV Video Quality Assessment Model Based on Neural Network.