Voice and Video QoE Inference Beyond Static MOS Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring user experience in online media applications, such as video conferencing, rely on static formulas and mean opinion score (MOS) values that fail to accurately reflect the true quality of experience (QoE) due to their deterministic nature and inability to consider user-specific and network effects.
Innovation Solution
Utilize perception models to analyze media data and compute performance measures, allowing for quantification of QoE and enabling proactive configuration changes to improve user experience by predicting and mitigating potential SLA violations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If static formulas and MOS values are used to measure QoE, then the measurement method is simple and deterministic, but the accuracy of QoE measurement deteriorates because it cannot reflect true user experience and ignores user-specific and network effects
Solution Approach 1:
The patent replaces traditional mechanical/statistical QoE measurement methods (MOS formulas) with machine learning perception models. These models use neural networks to analyze media data and predict user experience metrics, substituting deterministic mathematical formulas with adaptive intelligent systems that can capture complex non-linear relationships between network parameters and actual user perception.
Solution Approach 2:
The patent transforms QoE measurement from using fixed network parameters (loss, latency, jitter) through static formulas to using dynamic perception results from ML models. The system changes the measurement parameters by incorporating media quality assessments, user-specific characteristics, and application-level metrics that adapt to different conditions, thereby improving measurement accuracy while managing complexity through modular model design.
2Measurement precision
If perception models are used to analyze media data and compute performance measures, then the accuracy of QoE measurement is improved by considering dynamic user and network factors, but the device complexity increases
Solution Approach 1:
The patent segments the QoE measurement system into distinct perception models for different media types (audio, video) and different QoE dimensions. Each model is specialized and can be independently trained and deployed, allowing the system to manage complexity through modular architecture while achieving high measurement accuracy through dedicated models for specific tasks.
Solution Approach 2:
The patent introduces perception models as intermediary components between raw media data and QoE metrics. These models act as intelligent mediators that process complex media inputs and translate them into meaningful QoE measurements, bridging the gap between raw data and actionable insights while managing system complexity through standardized model interfaces.
3Ease of operation
If MOS values are used to indicate QoE, then the measurement method is simple to implement, but the ability to discern differences between media sessions deteriorates because MOS can only indicate failure and requires significant distribution differences to show variation
Solution Approach 1:
The patent transitions from static MOS scoring to dynamic perception-based QoE measurement. The perception models continuously adapt to different media sessions, user contexts, and network conditions, providing real-time differentiated assessments that capture subtle quality variations. This dynamic approach maintains ease of operation through automated model inference while dramatically improving the ability to distinguish between different session qualities.
Data Source
AI summary
In one embodiment, a device obtains perception results generated by one or more perception models that use media data as input that is transmitted between endpoints of an online application via a network. The device computes performance measures for the one or more perception models, based in part on the perception results and on the media data. The device quantifies, based on the performance measures, quality of experience for the online application. The device causes a configuration change to be made with respect to the online application, based on the quality of experience.