Audio Fingerprint Bucket Filtering for Faster Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audio fingerprinting systems face significant storage and resource challenges due to a subset of very common fingerprints that occur frequently across audio tracks, leading to inefficient use of storage space, increased processing requirements, and higher costs, without contributing significantly to discriminative power in matching processes.
Innovation Solution
Implementing optimization techniques such as overwriting and dropping buckets with high entry counts, maintaining count-based metrics, and selectively using discriminative fingerprints for matching, while utilizing secondary storage for non-discriminative data, to manage resource usage efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all fingerprints are stored in the database to ensure complete matching accuracy, then matching accuracy is maintained, but storage space and memory usage increase significantly
Solution Approach 1:
The patent extracts and separates common fingerprints that appear in many audio tracks from the main fingerprint database. These common fingerprints are stored separately or omitted from the primary storage structure, reducing the overall storage requirements while maintaining matching accuracy through selective retrieval strategies.
Solution Approach 2:
The patent applies different storage and retrieval strategies to different types of fingerprints based on their discriminative power. Rare, highly discriminative fingerprints are stored with high priority in the main database, while common, less discriminative fingerprints are handled differently (separated, compressed, or stored with lower priority), optimizing resource allocation according to local needs.
2Speed
If all fingerprints are kept in memory for fast access, then query speed is improved, but memory consumption increases
Solution Approach 1:
The patent segments the fingerprint database into different parts: frequently accessed, highly discriminative fingerprints are kept in memory for fast access, while less frequently accessed or less discriminative fingerprints are stored on disk or in secondary storage. This segmentation allows the system to maintain fast query performance for critical fingerprints while reducing overall memory consumption.
Solution Approach 2:
The patent implements dynamic memory management where the contents of memory are adaptively adjusted based on query patterns and system resources. Frequently accessed fingerprints are loaded into memory, while less accessed ones are kept on disk, with automatic eviction and reloading based on usage patterns, optimizing the balance between speed and memory usage.
3Reliability
If common fingerprints are stored to maintain completeness, then database completeness is preserved, but processing time increases
Solution Approach 1:
The patent extracts common fingerprints that appear in many audio tracks and handles them separately from the main matching process. These common fingerprints are either pre-processed, stored in a specialized structure, or retrieved on-demand, reducing the processing time for the main matching operations while preserving database completeness through selective inclusion.
4Productivity
If the fingerprint database is optimized to reduce storage, then resource efficiency is improved, but match accuracy may degrade
Solution Approach 1:
The patent optimizes storage by applying different compression and storage strategies to different types of fingerprints based on their discriminative power. Highly discriminative, rare fingerprints are stored with high fidelity and priority, while common, less discriminative fingerprints are compressed or stored with lower priority, achieving resource efficiency without significantly degrading overall match accuracy.
Data Source
AI summary
Techniques are disclosed to optimize resource usage and matching speed of audio track fingerprinting and matching systems without causing excessive impacts on match accuracy. Fingerprints of corpus of audio tracks are recorded via a fingerprint table for use to match query tracks with known tracks in the corpus. However, very large buckets in the fingerprint table create strain on system resources and typically add little to no discriminative power to improve track matching accuracy. Thus, in embodiments, the disclosed system tracks the entry count of individual buckets, and drops bucket data of a bucket when the bucket reaches a count threshold. In embodiments, the disclosed system performs another technique during the matching process to exclude the highest-count fingerprint buckets from the matching process. These techniques may be adjusted according to user configuration or empirical data to maintain a high match accuracy level despite using fewer fingerprints.