Speech Recognition TRIZ Cases
- Dynamic NLU Execution with Caching to Reduce Latency and Cost
A natural language understanding system uses dynamic execution delays and intermediate result caching to reduce processing latency.
- Voice Emergency Stop Control Device for Robot Safety
A control device detects voice segments to stop a robot using processor analysis of sound data.
- Sign Language Interpretation Overlay for Media Accessibility
A receiving device converts closed captioning text into sign language video overlays for simultaneous display alongside media content.
- Voice Input System for Vehicle Message Routing
A voice input system distinguishes human and machine intent to route messages automatically.
- Voice Control System for Remote Device Accessibility
A voice control system processes acoustic signals to operate target devices remotely without portable hardware.
- Speech Recognition Candidate Selection via Dynamic Rule Execution
A speech recognition system executes selection rules with varying complexity to identify target results from candidate outputs.
- Speech Signal Processing via Subband Spectrum Envelope Prediction
Predicts reduced subband spectrum envelopes to adjust initial spectra, reducing quantization noise in real-time communication.
- Noise-Added VAD Model for Accurate Voice Signal Boundary Detection
A noise-added voice activity detection model classifies audio frames using deep neural networks to identify speech segments.
- Speech Signal Processing Adjusting Probability Values
A speech signal processing apparatus adjusts candidate output token probabilities based on priority levels to enhance recognition accuracy.
- Speech Recognition Arbitration Logic Using Categorical Confidence Levels
Categorical confidence levels preserve low-score data for user validation, improving task completion rates without numerical normalization.