Video Quality Analysis Using Content-Type Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing real-time video quality assessment tools are limited in their ability to analyze video quality due to overfitting and other issues, particularly in service provider systems where video degradation occurs, leading to inaccurate perception of video quality.
Innovation Solution
A method and system for analyzing video quality by sampling video at specific points in the service provider system, determining event features that characterize source events, and using a no-reference quality evaluator trained with these features to assess video quality across different delivery path points, providing a quality metric for each sampled video.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If generic video quality assessment models are used, then the system can process any video content, but the measurement precision deteriorates due to overfitting and inability to capture content-specific characteristics
Solution Approach 1:
The patent segments video content into distinct content types (e.g., sports, news, entertainment, weather) and creates separate quality assessment models for each type. This segmentation allows each model to specialize in specific content characteristics, improving measurement precision for each content type while maintaining overall system versatility through the collection of specialized models.
Solution Approach 2:
The patent changes the parameters and features used in quality assessment based on content type. Different content types have different important features (e.g., sports may prioritize motion clarity, while news may prioritize static image quality). By adapting parameters according to content type, the system achieves higher measurement precision without sacrificing adaptability.
2Productivity
If real-time video quality assessment is implemented, then the productivity increases, but the device complexity increases due to the need for multiple content-specific models and processing infrastructure
Solution Approach 1:
The patent performs preliminary classification of video content into content types before applying specific quality assessment models. This preliminary action simplifies the overall system complexity by routing videos to appropriate pre-trained models, avoiding the need for a single complex universal model while maintaining real-time processing capability.
Solution Approach 2:
The patent creates a universal framework that can handle multiple content types through a common architecture with specialized components. The system uses a unified processing pipeline that adapts to different content types, reducing device complexity compared to having completely separate systems for each content type while still enabling real-time multi-format assessment.
3Measurement precision
If video is sampled at multiple delivery path points, then the measurement precision improves for identifying degradation sources, but the loss of time increases due to multiple sampling operations
Solution Approach 1:
The patent strategically skips certain sampling points or reduces sampling frequency at points where degradation is less likely to occur or where the impact on measurement precision would be minimal. This allows the system to maintain high degradation source identification accuracy while reducing the total time spent on multiple sampling operations.
Solution Approach 2:
The patent introduces an intermediary rapid analysis method that performs quick quality checks at intermediate sampling points. This intermediary approach provides sufficient information to identify major degradation sources without requiring full-depth analysis at every point, thus reducing time loss while maintaining measurement precision for critical issues.
Data Source
AI summary
Methods and systems for analyzing video quality of programming by content type. A method for analyzing video quality includes receiving, at a service provider system, video containing a source event. The system can sample the video at a least processed sampling point in the service provider system. An event feature set characterizing the source event is determined using the sampled video. The event feature set includes unique features that are same and unique features that are substantially same in videos including future occurrences of the source event. The system assesses the quality of one or more sampled videos taken at other delivery path points in the service provider system. The quality assessed using a source event no reference quality evaluator trained using at least the event feature set. The system outputs a quality metric for each of the one or more sampled videos taken at the other delivery path points.