Ontology-Aware Sound Classification via Dual Graph Convolutional Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audio classification systems lack the ability to effectively categorize ambiguous and overlapping sounds, failing to disambiguate acoustically similar yet semantically different sounds, and struggle with multi-label classification, which limits their ability to understand the context of sound origins.
Innovation Solution
The implementation of an ontology-aware sound classification system using graph convolutional networks to create relationships between fine and coarse audio classification labels, allowing for the classification of sounds by leveraging the likelihood of co-occurrence in real-world scenarios, thereby enhancing the system's ability to distinguish between similar sounds and classify them within general and specific categories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional audio classification systems are used, then the system structure is simple, but the system cannot disambiguate acoustically similar yet semantically different sounds
Solution Approach 1:
The audio classification task is segmented into two distinct levels: fine-grained classification (specific sound types) and coarse-grained classification (broad sound categories). This is achieved by creating separate graph structures for fine labels and coarse labels, allowing the system to process different classification granularities independently while maintaining overall accuracy.
Solution Approach 2:
The patent introduces an ontological dimension to the classification system by organizing sound labels into a hierarchical structure with fine labels (specific sounds) and coarse labels (broad categories). This dimensional organization allows the system to navigate between specific and general classifications, resolving ambiguities that single-level systems cannot handle.
2Adaptability or versatility
If traditional audio classification systems are used, then the processing speed is fast, but the system cannot classify overlapping sounds in multi-label scenarios
Solution Approach 1:
The patent introduces an ontology-based graph structure as an intermediary that captures co-occurrence relationships between different sound labels. This graph acts as a mediator that provides contextual information about how sounds typically appear together in real-world scenarios, enabling the system to resolve overlapping sounds more effectively.
Solution Approach 2:
The system pre-computes and stores co-occurrence statistics and ontological relationships between sound labels in graph structures before actual classification. This preliminary action allows the system to quickly query and utilize contextual relationships during inference without performing complex computations in real-time, maintaining processing speed while improving multi-label classification capability.
3Measurement precision
If audio classification systems utilize ontology, then the system can disambiguate similar sounds, but the system complexity increases due to graph construction and processing
Solution Approach 1:
The patent creates simplified graph representations that copy only the essential co-occurrence and hierarchical relationships from the full ontology. Rather than processing the complete ontological structure, the system uses condensed graph versions that retain the critical information needed for disambiguation while reducing computational complexity.
Data Source
AI summary
A system for ontology-aware sound classification. The system includes an electronic processor that is configured to create a first graph based on relationships between fine audio classification labels and create a second graph based on relationships between coarse audio classification labels. The electronic processor is also configured to receive an audio clip including one or more sounds, execute a first graph convolutional network with the first graph as input, and execute a second graph convolutional network with the second graph as input. Using the outputs of the first graph convolutional network and the second graph convolutional network, the electronic processor is configured to determine one or more coarse labels, one or more fine labels, or both to classify the one or more sounds in the audio clip.