Research On User Perception Quality Of Audio And Video Services In Wireless Network

The rapid development and the performance promotion of wireless network enrich the service form greatly, wherein, the audio service and the video service have become the most important service types in the current wireless network. While the assessment standards for the quality of existing services are based on the Quality of Service (QoS) of the network and devices, which are failed to be in conformity with the end user’s perception completely. Therefore, Quality of Experience (QoE) is introduced as a new dimension to assess the quality of audio service and video service to the wireless network. However, the subjective QoE assessment method requires plenty of manpower and a lot of time, and is a posteriori for the assessment result of a specific service. Thus, it is how to accurately and effectively assess the perceptual quality of the specific service with a priori objective method becomes a hot research topic because of its important theoretical and practical significance. Meanwhile, as one of the service quality standards, it is no enough for QoE to just provide the quality result of a service, the source of service perceptual quality impact should be located and quantized. Under this kind of background, the research aimed to different links of video service quality impact is also of great necessity.The thesis mainly researches the objective QoE assessment method of the audio service and the video service, establishes multiple QoE models aimed to the specific scenes. The performance of Perceptual Evaluation of Speech Quality (PESQ) in the ITU P.862 Standard is analyzed and verified pointed to the Chinese audio service. For the video service, the thesis establishes a QoE model for the end to end service, researches the perceptual quality impact influences of video service caused by encoding compress and network transmission respectively, and finally establishes a QoE model of a novel adaptive streaming service.The main contributes of the thesis includes:1. Build a Chinese speech corpus in strict conformance with the requirement of ITU Standard. Establish transmission impact models and simulate the speech service transmission. Obtain speech samples with real transmission impact, and get the perceptual quality of the Chinese speech corpus via plenty of subjective tests. Based on which, analyze the PESQ performance in Chinese speech environment, and then propose a verified model CPESQ (Chinese PESQ) to assess the Chinese speech perceptual quality. The analysis results show that, the CPESQ model adapts to the subjective perceptual quality better, and promotes the correlation and the accuracy, compared with PESQ.2. In the end to end video service, the influence factors that impact the video service quality are abstracted. The influences to the perceptual quality caused by service indicators (initial delay, play stall, synchronization of video and audio) and image indicators (block, blur, motion) are analyzed. Then the thesis propose a video service QoE model VQAM (Video Quality Assessment Model) is proposed by regression analyzing the relation among the above factors.3. Aimed to the video encoding compress, a no reference video perceptual quality assessment model, NR-VMOS (No Reference-Video MOS), is proposed. The model introduces the no reference PSNR estimation and establishes different sub-model aimed to different encoding size and compressed code rate. The analysis results show that, the above model has great adaptation to the videos with different contents, and can be implemented on the video service platform to assess the performance of different encoders.4. Research the influence to the perceptual quality of the streaming service caused by the network transmission. Discuss the influence to the streaming services with different transmission manners:1) Research the image quality degradation in the streaming service caused by the transmission packet loss by introducing SSIM (Structural SIMilarity) algorithm; analyze the error from the packet loss and the error propagation in Inter-frame Prediction codec; Propose and analyze two e-SSIM estimation methods caused by network packet loss, based on which, establishes a streaming service perceptual quality model of packet loss.2) Research the stall influence caused by delay and the insufficient throughput. Propose an analytical model of stall distribution and transmission indicators:packet loss, delay and throughput, and the model is validated through simulation. Analyze the relationship between subjective perceptual quality and stall factors based on psychology.5. Discuss the perceptual quality of adaptive streaming service, which contains the encoding impact fluctuation and play stall impact. The thesis illustrates the influence of the perceptual quality caused by bitrate change, bitrate distribution, stall, initial bitrate. And based on initial bitrate, stall and bitrate change, establishes a QoE model of DASH (Dynamic Adaptive Streaming over HTTP) service. Perform the regression analysis through plenty of crowd tests, and by comparing with other similar model, the proposed model can derivate the real-time perceptual quality of DASH service accurately.