Reliable and repeatable video quality assessment is essential for performance analysis of wireless multimedia applications in the third generation (3G) Long Term Evolution (LTE) network. In this article we report on a database containing subjective assessment scores and the corresponding Quality-of-Service parameters of 70 video test sequences encoded with H.264, which are corrupted when transmitted over a wireless 3G LTE network simulator. Then, a new assessment method based on neural networks (NN) is proposed, whose weights are determined through training.
The resulting pseudo-subjective assessment scores are then compared to the true MOS results in our database. Naturally, the accuracy of the NN-based prediction tool should be tested `outside’ of the set used for NN weight training. However, there are persistent residual errors between the predicted and subjectively evaluated MOS, which can be further reduced by particle swarm optimization, applied as post-processing of the NN weights to improve its accuracy. The proposed assessment method has potential applications, including Quality-of-Experience-aware network optimization for LTE network operators.