A real-time blind quality-of-experience assessment metric for HTTP adaptive streaming

Chunyi Li,May Lim,Abdelhak Bentaleb,Roger Zimmermann

2023-03-17

Abstract:In today's Internet, HTTP Adaptive Streaming (HAS) is the mainstream standard for video streaming, which switches the bitrate of the video content based on an Adaptive BitRate (ABR) algorithm. An effective Quality of Experience (QoE) assessment metric can provide crucial feedback to an ABR algorithm. However, predicting such real-time QoE on the client side is challenging. The QoE prediction requires high consistency with the Human Visual System (HVS), low latency, and blind assessment, which are difficult to realize together. To address this challenge, we analyzed various characteristics of HAS systems and propose a non-uniform sampling metric to reduce time complexity. Furthermore, we design an effective QoE metric that integrates resolution and rebuffering time as the Quality of Service (QoS), as well as spatiotemporal output from a deep neural network and specific switching events as content information. These reward and penalty features are regressed into quality scores with a Support Vector Regression (SVR) model. Experimental results show that the accuracy of our metric outperforms the mainstream blind QoE metrics by 0.3, and its computing time is only 60\% of the video playback, indicating that the proposed metric is capable of providing real-time guidance to ABR algorithms and improving the overall performance of HAS.

Image and Video Processing,Multimedia

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to predict the user's Quality of Experience (QoE) in real - time at the client side in the HTTP Adaptive Streaming (HAS) system. Specifically, the goal of the paper is to design a QoE metric method for blind evaluation (i.e., without the need for the original video reference), which can:

1. **High consistency with the Human Visual System (HVS)**: Ensure that the QoE metric is highly consistent with human perception of video playback quality.
2. **Low latency**: Provide feedback in real - time during video playback, rather than evaluating after transmission, so its calculation time cannot exceed the playback time of the video segment.
3. **Blind evaluation**: Perform real - time prediction at the client side, using only compressed or distorted videos, without relying on the original uncompressed source video or reference features obtained separately from the server.

To achieve these goals, the authors propose a new QoE metric method. This method reduces the time complexity through non - uniform sampling, combines resolution and re - buffering time as Quality of Service (QoS) indicators, and uses spatio - temporal outputs from deep neural networks and specific switching events as content information. Finally, these reward and penalty features are regressed into quality scores through the Support Vector Regression (SVR) model. Experimental results show that the performance of this method in two databases is better than that of mainstream blind QoE metric methods, and the calculation time is only 60% of the video playback time, meeting the needs of real - time guiding the ABR algorithm and improving the overall performance of HAS.