Sound Primitive Recognition via Finite-State Automata
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated systems are unable to detect higher-level events associated with collections of related sounds, as they struggle to recognize sequences or simultaneous occurrences of these sounds.
Innovation Solution
A system that uses a non-deterministic finite-state automaton to recognize sequences of sound primitives, feeding them through multiple levels of finite-state automata to generate recognized events, and associates semantic labels with sound primitives through an unsupervised learning process, including feature detection, clustering, and user feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing automated systems are used for sound recognition, then basic sound detection is achieved, but higher-level events associated with collections of related sounds cannot be detected
Solution Approach 1:
The system segments sound recognition into hierarchical levels: sound primitives (basic acoustic units) are detected first, then combined into sound events through finite-state automata. This segmentation allows the system to build complex event detection capabilities from simpler primitive recognition, resolving the contradiction between versatility and complexity.
Solution Approach 2:
The patent implements nested hierarchical structure where sound primitives are nested within sound events, which are nested within higher-level activities. Multiple levels of finite-state automata process events at different hierarchical levels, enabling detection of complex higher-level events while maintaining manageable system complexity through organized nesting.
2Adaptability or versatility
If systems attempt to recognize sequences of related sounds, then higher-level event detection is enabled, but the complexity of handling different sequences and simultaneous occurrences increases
Solution Approach 1:
The system uses dynamic finite-state automata that can adaptively process different sound sequences and simultaneous occurrences. The automata transition between states based on incoming sound primitives, dynamically handling various temporal patterns and sequences without requiring static, pre-programmed rules for every possible scenario.
Solution Approach 2:
Sound primitives serve as intermediaries between raw audio signals and higher-level event recognition. The finite-state automata act as intermediary processing layers that transform sequences of primitives into recognized events, simplifying the complexity of directly analyzing raw audio for complex patterns.
3Measurement precision
If feature detection and clustering operations are performed, then sound primitives are generated, but computational processing time increases
Solution Approach 1:
The system performs preliminary feature detection and clustering operations during an offline training phase to generate sound primitives and their associated acoustic models. This preliminary action allows the computationally intensive processing to be done in advance, so that during real-time operation, the system only needs to match incoming audio against pre-computed models, significantly reducing processing time while maintaining accuracy.
Data Source
AI summary
The disclosed embodiments provide a system that performs a sound-recognition operation. During operation, the system recognizes a sequence of sound primitives in an audio stream, wherein a sound primitive is associated with a semantic label comprising one or more words that describe a sound characterized by the sound primitive. Next, the system feeds the sequence of sound primitives into a finite-state automaton that recognizes events associated with sequences of sound primitives. Finally, the system feeds the recognized events into an output system that generates an output associated with the recognized events to be displayed to a user.