A study of real-time packet video quality using random neural networks

Abstract

An important and unsolved problem today is that of automatic quantification of the quality of video flows transmitted over packet networks. In particular, the ability to perform this task in real time (typically for streams sent themselves in real time) is especially interesting. The problem is still unsolved because there are many parameters affecting video quality, and their combined effect is not well identified and understood. Among these parameters, we have the source bit rate, the encoded frame type, the frame rate at the source, the packet loss rate in the network, etc. Only subjective evaluations give good results but, by definition, they are not automatic. We have previously explored the possibility of using artificial neural networks (NNs) to automatically quantify the quality of video flows and we showed that they can give results well correlated with human perception. In this paper, our goal is twofold. First, we report on a significant enhancement of our method by means of a new neural approach, the random NN model, and its learning algorithm, both of which offer better performances for our application. Second, we follow our approach to study and analyze the behavior of video quality for wide range variations of a set of selected parameters. This may help in developing control mechanisms in order to deliver the best possible video quality given the current network situation, and in better understanding of QoS aspects in multimedia engineering.Index Terms-Packet video, random neural networks, real-time video transmission, video quality assessment, video signal characterization.

No-Reference Quality Assessment for Networked Video via Primary Analysis of Bit Stream

“…In [16], a simple method was proposed to assess the quality of the networked video relative to bit-rate and packet loss rate. Another approach based on an artificial neural network [17] utilized, in addition to bit-rate and packet loss rate, other information including the frame rate, the number of consecutively lost packets, and the I-block refresh rate. It is widely acknowledged that the impact of packet loss on the perceived video quality is dependent on video content and the position of the lost packet in the video stream.…”

“…Various techniques have been reported in the literature for estimating video quality relative to traditional objective metrics, e.g., the peak signal-tonoise ratio (PSNR) [11]- [14]. NR perceptual metrics have also been reported which compute perceived video quality based on coding rate and packet loss rate [16], or the use of an artificial neural network [17]. For NR stream analysis methods, the video quality prediction performance highly depends on the level of access to the video stream [18].…”

Optimal Quality-of-Experience Design for a P2P Multi-Source Video Streaming

“…PSQA [5] is a technique based on merging subjective assessments with a statistical learning tool (a Random Neural Network, or RNN, which allows to produce subjective-like quality estimations. Its main advantage is that it provides results very close to actual MOS values while being cheap and suitable for real-time applications.…”

“…First of all, we analyze the impact of frame losses on the perceived quality by the end-user (previous works considered losses at the packet level). We address the problem of measuring the perceived quality by means of the PSQA (Pseudo Subjective Quality Assessment) technology, improving from [5]. Second, we analyze the behavior of the system when peers leave.…”

A robust P2P streaming architecture and its application to a high quality live-video service

“…To evaluate the quality at the client side, we use the PSQA technology. Pseudo-Subjective Quality Assessment (PSQA) [3] is a general procedure that automatically measures the perceived quality, accurately and in real time. For instance, if we assume that the PSQA metric is scaled on [0..1] and if the network is perfect, its instantaneous total quality is equal to the number of connected clients (because P SQA = 1 for each client).…”