Sound Data Pattern Matching Using Hashing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pattern matching techniques for sound data are resource-intensive, limiting their effectiveness in real-time scenarios such as interactive systems and making them unsuitable for applications like audio source separation and word spotting.
Innovation Solution
The use of hashing techniques in conjunction with nonnegative matrix factorization (NMF) to efficiently represent and process sound data, allowing for dynamic selection of relevant entries and updates of hash codes during the NMF process, thereby reducing computational complexity and enabling real-time pattern matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional pattern matching techniques are used for sound data, then pattern matching functionality is achieved, but computational resource consumption increases and real-time performance deteriorates
Solution Approach 1:
The patent creates a compressed representation (hash code) of the spectrogram that captures essential acoustic characteristics. This hash code serves as a simplified copy that can be quickly compared against database entries without processing the full spectrogram data, thereby reducing computational resources while maintaining pattern matching capability
Solution Approach 2:
The patent extracts key acoustic features from the full spectrogram by applying hashing operations that identify and retain only the most relevant spectral characteristics. This extraction process discards redundant information while preserving the essential patterns needed for sound data matching, significantly reducing the computational burden
2Measurement precision
If large dictionaries are used for pattern matching, then matching accuracy improves, but processing time increases and real-time capability is lost
Solution Approach 1:
Instead of storing and comparing full spectrograms in the database, the patent stores compressed hash codes that represent the essential acoustic patterns. This allows the system to maintain a comprehensive dictionary of sound patterns while enabling rapid comparison through simple hash code matching, thus preserving accuracy while dramatically reducing processing time
Solution Approach 2:
The patent pre-computes and stores hash codes of reference sound patterns in the database during an offline training phase. During real-time operation, the system only needs to compute the hash code of the input spectrogram and compare it against the pre-stored hash codes, eliminating the need for time-consuming full spectrogram comparisons during critical real-time operations
Data Source
AI summary
Pattern matching of sound data using hashing is described. In one or more implementations, a query formed from one or more spectrograms of sound data is hashed and used to locate one or more labels in a database of sound signals. Each of the labels is located using a hash of an entry in the database. At least one of the located one or more labels is chosen as corresponding to the query.