Multi-pass fingerprint indexing with subsampling and parallel processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying media content through fingerprint matching, such as brute force comparisons and lookup tables, are inefficient due to high computational costs and bit error rates, especially when dealing with large databases of media content.
Innovation Solution
The proposed solution involves a multi-pass approach with subsampling and parallel processing, using GPUs to compare query fingerprints against down-sampled reference fingerprints, and employing bit error rate optimization to reduce the number of comparisons needed, thereby improving matching efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If brute force comparison methods are used for fingerprint matching, then comprehensive search coverage is achieved, but computational cost and processing time increase significantly
Solution Approach 1:
The patent segments the fingerprint matching process into multiple passes: a first pass using down-sampled fingerprints for rapid filtering, and a second pass using full-resolution fingerprints for precise matching. This segmentation allows comprehensive search coverage while dramatically reducing overall processing time by eliminating the need to perform full comparisons on all database entries.
Solution Approach 2:
The patent performs preliminary down-sampling of fingerprints to create a reduced representation before the main matching process. This preliminary action creates a smaller search space that can be quickly scanned, allowing the system to identify candidate matches without immediately processing all full-resolution fingerprints, thus reducing computational cost while maintaining reliability.
2Measurement precision
If full-resolution fingerprints are compared against all reference fingerprints, then matching accuracy is maximized, but memory usage and computational resources increase
Solution Approach 1:
The patent applies local quality by using different resolution levels for different stages of the matching process. Down-sampled fingerprints are used for the initial filtering stage where high precision is not critical, while full-resolution fingerprints are reserved for the final matching stage where accuracy is paramount. This approach optimizes memory usage by storing only essential high-resolution data.
Solution Approach 2:
The patent creates a down-sampled copy of each fingerprint that serves as a simplified representation for rapid comparison. This copying strategy allows the system to work with smaller data structures during the filtering phase, reducing memory requirements, while the original full-resolution fingerprints remain available for accurate matching when needed.
3Productivity
If down-sampled fingerprints are used for initial filtering, then processing speed increases, but bit error rate increases
Solution Approach 1:
The patent performs preliminary filtering using down-sampled fingerprints to quickly eliminate non-matching candidates. This preliminary action with reduced precision is acceptable because it only needs to filter out obvious non-matches, not to make final determination. The bit errors introduced at this stage do not compromise overall reliability because subsequent full-resolution comparison validates all candidate matches.
Solution Approach 2:
The down-sampled fingerprint acts as an intermediary between the query and the full-resolution reference fingerprints. It performs the initial filtering function with reduced computational requirements, while the full-resolution fingerprints serve as the ultimate arbiter for accurate matching. This intermediary approach balances processing speed with reliability by using the appropriate level of detail for each stage.
Data Source
AI summary
Example methods and systems for indexing fingerprints are described. Fingerprints may be made up of sub-fingerprints, each of which corresponds to a frame of the media, which is a smaller unit of time than the fingerprint. In some example embodiments, multiple passes are performed. For example, a first pass may be performed that compares the sub-fingerprints of the query fingerprint with every thirty-second sub-fingerprint of the reference material to identify likely matches. In this example, a second pass is performed that compares the sub-fingerprints of the query fingerprint with every fourth sub-fingerprint of the likely matches to provide a greater degree of confidence. A third pass may be performed that uses every sub-fingerprint of the most likely matches, to help distinguish between similar references or to identify with greater precision the timing of the match. Each of these passes is amenable to parallelization.