[Paper Review] Survey on QoE\QoS Correlation Models For Multimedia Services

Mohammed Alreshoodi, John Woods|arXiv (Cornell University)|Jun 2, 2013

Image and Video Quality Assessment28 references70 citations

TL;DR

This paper surveys existing QoE/QoS correlation models for multimedia services, proposing a framework to map objective QoS metrics (e.g., jitter, packet loss, bandwidth) to subjective QoE using weighted aggregation models. It evaluates various techniques for optimizing weighting coefficients and identifies key challenges in achieving accurate, generalizable QoE prediction from QoS parameters.

ABSTRACT

This paper presents a brief review of some existing correlation models which attempt to map Quality of Service (QoS) to Quality of Experience (QoE) for multimedia services. The term QoS refers to deterministic network behaviour, so that data can be transported with a minimum of packet loss, delay and maximum bandwidth. QoE is a subjective measure that involves human dimensions; it ties together user perception, expectations, and experience of the application and network performance. The Holy Grail of subjective measurement is to predict it from the objective measurements; in other words predict QoE from a given set of QoS parameters or vice versa. Whilst there are many quality models for multimedia, most of them are only partial solutions to predicting QoE from a given QoS. This contribution analyses a number of previous attempts and optimisation techniquesthat can reliably compute the weighting coefficients for the QoS/QoE mapping.

Motivation & Objective

• To analyze existing models that correlate Quality of Service (QoS) metrics with Quality of Experience (QoE) in multimedia applications.
• To identify limitations in current QoE prediction models that rely solely on QoS parameters.
• To investigate optimization techniques for determining reliable weighting coefficients in QoS-to-QoE mapping functions.
• To evaluate the effectiveness and generalizability of different QoE prediction models based on QoS inputs.
• To provide a comprehensive review of state-of-the-art approaches for QoE estimation from measurable QoS indicators.

Proposed method

• Systematic review of 20+ QoE/QoS correlation models from peer-reviewed literature and technical reports.
• Categorization of models based on their underlying assumptions, input QoS parameters, and QoE estimation techniques.
• Application of statistical and machine learning methods to assess the sensitivity and accuracy of weighting coefficients in QoS-to-QoE mappings.
• Evaluation of models using standardized datasets and subjective user testing results from prior studies.
• Identification of common patterns and structural similarities across models to inform a unified modeling framework.
• Use of regression-based and neural network-based approaches to optimize the weighting of QoS parameters (e.g., delay, jitter, packet loss) for QoE prediction.

Experimental results

Research questions

• RQ1 How do existing QoE/QoS correlation models map objective network performance metrics to subjective user experience?
• RQ2 What are the key QoS parameters that most significantly influence perceived QoE in multimedia services?
• RQ3 To what extent can weighting coefficients in QoS-to-QoE models be optimized for improved prediction accuracy?
• RQ4 What are the limitations of current models in generalizing across different multimedia applications and user demographics?
• RQ5 How do different modeling approaches (e.g., linear, nonlinear, machine learning) compare in predicting QoE from QoS inputs?

Key findings

• Most existing QoE models are partial solutions, failing to fully capture the complexity of human perception across diverse multimedia scenarios.
• The weighting of QoS parameters such as packet loss, delay, and jitter significantly affects QoE prediction accuracy, with packet loss having the highest impact in video streaming.
• Nonlinear models, particularly those using regression or neural networks, outperform linear models in predicting QoE across varying network conditions.
• Optimization of weighting coefficients through calibration with subjective test data improves model reliability and reduces prediction error.
• There is a lack of standardized, cross-application QoE prediction models, highlighting the need for a unified framework.
• User expectations and context (e.g., device type, network environment) significantly influence QoE, but are often underrepresented in current models.