CONTINUOUS DETERMINATION OF QUALITY OF EXPERIENCE IN ENCRYPTED VIDEO TRAFFIC USING SEMI-SUPERVISED LEARNING WITH GENERATIVE ADVERSARIAL NETWORKS

摘要

A method and system for classifying features of an encrypted video stream to determine video quality of experience, QoE. At least one stream-related feature (170) is extracted from one or more encrypted video streams (130) sent from a content server (120) to at least one user device (110). and input into a discriminator neural network (150) of a generative adversarial network (210). The discriminator neural network is trained to classify the at least one stream-related feature using semi-supervised learning based on labeled data (250) and first unlabeled data (280). The discriminator neural network is periodically re-trained using the labeled data and second unlabeled data based at least in part on meeting a predetermined condition. The discriminator neural network classifies the at least one stream-related feature into at least one class of a plurality of classes, each of the plurality of classes corresponding to a QoE parameter (160).