Multimedia Session MOS Estimation via Segmented Buffer Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing buffer aggregation models are limited to session lengths of up to one minute, failing to accurately predict Mean Opinion Scores (MOS) for longer video sessions, which are common in streaming services like YouTube, and do not account for human memory effects and quality variations over time.
Innovation Solution
A method for predicting multimedia session MOS that generates audiovisual quality features using polynomial functions and exponential weights, incorporating initial buffering, rebuffering events, and oscillations, along with buffering features to estimate the overall MOS, capable of handling both short and long video sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing buffer aggregation models are used, then MOS estimation is simple and fast, but accuracy is poor for video sessions longer than one minute
Solution Approach 1:
The patent segments the video session into multiple one-minute intervals, each processed by the existing buffer aggregation model. This allows the simple model to be applied repeatedly to capture local quality variations while maintaining computational efficiency. The segmented approach enables accurate tracking of quality changes over longer durations without requiring a completely new complex model.
Solution Approach 2:
The patent introduces a temporal dimension by processing video sessions in sequential one-minute segments rather than treating them as a single unit. This dimensional transformation allows the application of simple aggregation models at each time point while capturing overall session quality through temporal progression, effectively extending the applicability of simple models to long sessions.
2Measurement precision
If subjective tests are used to evaluate quality, then MOS scores are accurate, but time and money consumption increase
Solution Approach 1:
The patent creates an objective copy of the subjective MOS evaluation process by training machine learning models on subjective test data. Once trained, these models can predict MOS scores without requiring actual human viewers, thus replicating the accuracy of subjective tests while eliminating the time and resource costs of conducting new subjective evaluations.
Solution Approach 2:
The patent transforms the evaluation process from direct human perception (subjective testing) to mathematical parameter analysis (objective modeling). By changing the evaluation parameters from human ratings to measurable video/audio parameters processed through trained models, the system achieves comparable accuracy without the temporal and resource constraints of subjective testing.
3Reliability
If video bitrate is reduced to adapt to network conditions, then video delivery is maintained, but quality decreases
Solution Approach 1:
The patent implements dynamic quality assessment by continuously evaluating video and audio parameters throughout the session. This allows the system to adaptively track quality changes as bitrate varies, providing accurate MOS scores that reflect the actual user experience at each moment rather than using static quality assessments.
Solution Approach 2:
The patent creates a feedback mechanism where quality parameters are continuously measured and fed into the aggregation model. This feedback loop enables the system to respond to quality variations caused by bitrate adaptation, maintaining reliable delivery assessment while accurately reflecting quality changes through real-time parameter monitoring.
Data Source
AI summary
Mechanisms for predicting a multimedia session MOS are provided. The multimedia session has a video session and an audio session. The video quality and the audio quality are represented by vectors of per-time-unit scores of video quality and audio quality, respectively. The multimedia session is represented by a vector of the rebuffering start times of each rebuffering event, and a vector of the rebuffering durations of each rebuffering event. Audiovisual quality features are generated from the vectors of per-time-unit scores of video and audio quality. Buffering features are generated from the vector of rebuffering start times of each rebuffering event and the vector of rebuffering durations of each rebuffering event. A multimedia session MOS is then estimated based on the generated audiovisual quality features and the generated buffering features.