Differential Scoring for Video Content Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video content matching systems struggle to accurately identify videos with similar content, often producing false negatives or false positives due to their inability to handle slight deviations in video files, particularly in cases like sports videos with similar visual content.
Innovation Solution
A method that divides input videos into sequential segments, identifies similar segment groups in candidate videos, and evaluates similarity measures over time intervals to determine the likelihood of matching video content, providing an indication when the likelihood exceeds a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If hash-based fingerprint functions are used to identify exact copies of video files, then identification precision for exact matches is improved, but tolerance for slight deviations between files deteriorates, producing false negative results
Solution Approach 1:
The patent segments video files into multiple blocks and generates fingerprints for each block separately, then combines them. This segmentation allows the system to identify videos even when only portions match, thereby tolerating slight deviations while maintaining identification precision for matching segments.
Solution Approach 2:
The patent changes the fingerprinting approach from monolithic to multi-block with configurable parameters such as block size, overlap regions, and similarity thresholds. These parameter adjustments enable the system to balance between exact match precision and tolerance for deviations by tuning the sensitivity of each fingerprint block comparison.
2Reliability
If more tolerant identification approaches are used to handle deviations between files, then reliability for slight deviations is improved, but identification precision deteriorates, producing false positive identifications
Solution Approach 1:
By dividing the video into multiple blocks and requiring a threshold number of matching blocks rather than requiring all blocks to match, the system achieves tolerance for deviations without sacrificing precision. The segmentation allows selective matching of relevant portions while ignoring non-critical differences.
Solution Approach 2:
The patent implements partial matching where only a portion of the video blocks need to match to identify similar videos. This partial action approach provides tolerance for deviations in non-matching blocks while maintaining precision through the requirement that sufficient blocks must match the similarity threshold.
3Measurement precision
If manual identification by humans is used to accurately identify video content, then identification precision is improved, but productivity and cost efficiency deteriorate
Solution Approach 1:
The patent replaces manual human identification with an automated computer-based fingerprinting system that processes video files algorithmically. This substitution maintains high identification precision through sophisticated block-based comparison while dramatically improving productivity by automating the entire identification process without human intervention.
4Productivity
If simple fingerprint comparison is used to quickly identify video matches, then productivity is improved, but measurement precision deteriorates, unable to distinguish videos with similar content
Solution Approach 1:
The patent segments videos into blocks and generates fingerprints for each block, enabling parallel processing that maintains speed while improving precision. The segmented approach allows the system to compare specific content regions independently, providing more precise content matching information than simple monolithic fingerprint comparison while maintaining computational efficiency.
Solution Approach 2:
The system performs partial fingerprint comparisons on selected blocks rather than requiring complete video comparison. This partial action provides sufficient precision for content matching by focusing on key segments while maintaining high productivity through reduced computational overhead compared to full video analysis.
Data Source
AI summary
Whether a not a given video includes video content of other videos is determined by evaluating video content as well as video content transitions over time of videos. One or more fingerprints of the video are created and used to match against reference fingerprints of reference videos to identify candidate videos that is likely to have matching video content as an input video. After a candidate video is identified, a likelihood of the candidate video and an input video having matching video content is evaluated. A confidence score based on the evaluation of video content as well as video content transitions over time for a pair of videos including the input video and a candidate video is generated. When the confidence score exceeds a threshold score, it is determined that the pair of videos is likely to include the matching video content.