Music Fingerprinting via Onset Interval Codes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio fingerprinting technologies rely heavily on spectral features, which are susceptible to strong stationary resonances and environmental noise, making them less effective in identifying music samples accurately across different recordings and formats.
Innovation Solution
A process that generates a fingerprint based on the relative timing of onsets within a music sample, using whitening, time-varying filters, and band-pass filtering to suppress resonances and noise, while detecting onsets and inter-onset intervals to create a robust code for identification, which can be used to search a music library database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If spectral features are used for audio fingerprinting, then the system can identify music samples, but the identification accuracy degrades due to strong stationary resonances and environmental noise
Solution Approach 1:
The patent extracts only the temporal onset information from the audio signal, separating it from the problematic spectral features. By focusing exclusively on onset timestamps and intervals rather than full spectral analysis, the system eliminates susceptibility to stationary resonances and environmental noise while maintaining identification capability
Solution Approach 2:
The patent replaces the traditional spectral-based fingerprinting mechanism with a temporal-based mechanism. Instead of analyzing frequency domain characteristics that are vulnerable to noise, the system uses time-domain onset detection and interval measurement, fundamentally substituting the feature extraction approach to achieve robustness
2Reliability
If traditional fingerprinting methods are used, then music samples can be identified, but the system performance degrades when sample lengths exceed presumed maximums
Solution Approach 1:
The patent segments the audio analysis into discrete onset events with timestamp and interval features. By representing the music sample as a sequence of onsets rather than continuous spectral analysis, the system can handle arbitrarily long samples without performance degradation, as each onset is independently detected and encoded
Solution Approach 2:
The patent employs dynamic thresholding and adaptive parameters for onset detection that adjust based on the local signal characteristics. This dynamic approach allows the system to maintain accuracy across varying sample lengths and conditions, rather than relying on fixed parameters that degrade with length
3Adaptability or versatility
If spectral features are used for fingerprinting, then the system can process music samples, but the robustness decreases across different recordings and formats
Solution Approach 1:
The patent creates a universal fingerprinting approach based on temporal onsets that functions across different recordings, formats, and conditions. By using onset timestamps and intervals as the universal feature set, the system achieves both adaptability to various inputs and robustness in identification, eliminating the trade-off present in spectral methods
Data Source
AI summary
Methods, computing devices, and machine readable storage media for generating a fingerprint of a music sample. The music sample may be filtered into a plurality of frequency bands. Onsets in each of the frequency bands may be independently detected. Inter-onset intervals between pairs of onsets within the same frequency band may be determined. At least one code associated with each onset may be generated, each code comprising a frequency band identifier identifying a frequency band in which the associated onset occurred and one or more inter-onset intervals. Each code may be associated with a timestamp indicating when the associated onset occurred within the music sample. All generated codes and the associated timestamps may be combined to form the fingerprint.