Voice Control Acoustic Recognition Parallel Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems for consumer electronics face challenges in achieving a balance between false reject and false accept rates, especially in noisy environments, and require manual triggers or internet access, limiting their hands-free functionality.
Innovation Solution
The system processes acoustic input signals using multiple recognition processes that start at different time points, with scores calculated across various states and offsets to improve recognition accuracy, allowing for hands-free operation and accurate word spotting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech recognition systems use a single recognition process with a fixed model, then the system complexity is low, but the recognition accuracy in noisy environments deteriorates
Solution Approach 1:
The patent divides the speech recognition process into multiple parallel recognition processes, each handling different segments or aspects of the acoustic input signal. Multiple acoustic models process the same input independently, and their results are combined to improve overall recognition accuracy while managing system complexity through modular design
Solution Approach 2:
The patent combines results from multiple independent recognition processes through score integration. Each process generates confidence scores for recognized words, and these scores are merged to produce a final recognition result with higher accuracy than any single process could achieve alone
2Ease of operation
If speech recognition systems process all acoustic input continuously, then hands-free operation is achieved, but the false accept rate increases in noisy environments
Solution Approach 1:
The patent implements a preliminary voice trigger mechanism that activates the full recognition system only when a specific trigger word is detected. This preliminary action filters out ambient noise by requiring a deliberate activation command, reducing false accepts while maintaining hands-free operation capability
Solution Approach 2:
The system dynamically adjusts its operation mode between trigger-based activation and continuous listening. The recognition process can switch between being dormant, actively listening for triggers, and fully engaged in command recognition, optimizing reliability based on operational context
3Measurement precision
If speech recognition systems use complex acoustic models to improve accuracy, then word spotting precision improves, but the processing time increases
Solution Approach 1:
The patent applies partial processing by focusing recognition efforts on specific time segments where trigger words are likely to occur. Instead of uniformly processing the entire acoustic signal with full complexity, the system applies intensive processing only to relevant segments, reducing overall processing time while maintaining precision
Data Source
AI summary
In one embodiment the present invention includes a method comprising receiving an acoustic input signal and processing the acoustic input signal with a plurality of acoustic recognition processes configured to recognize the same target sound. Different acoustic recognition processes start processing different segments of the acoustic input signal at different time points in the acoustic input signal. In one embodiment, initial states in the recognition processes may be configured on each time step.