Speech Recognition SNR Selection Across Multiple Apparatuses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition technologies face challenges in achieving optimal performance when users are distant from the recognition apparatus or in noisy environments, leading to reduced accuracy.
Innovation Solution
A method and apparatus that calculate signal-to-noise ratios (SNRs) of speech signals, determining a reference speech signal with the maximum SNR among multiple speech recognition apparatuses, and either recognizing it locally or transmitting it to another apparatus based on available resources, ensuring accurate speech recognition even in resource-constrained conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If speech recognition is performed in a noisy environment or when the user is distant from the apparatus, then the speech recognition accuracy deteriorates
Solution Approach 1:
The system segments the speech recognition task across multiple apparatuses by calculating SNR for each apparatus and selecting the reference signal from the apparatus with the maximum SNR. This segmentation allows the system to identify which apparatus has the cleanest signal and use that as the reference, thereby maintaining accuracy despite noisy environments or distance.
Solution Approach 2:
The system uses SNR calculation as feedback to determine which speech signal should be used as reference. By continuously monitoring the quality of speech signals through SNR computation, the system can adaptively select the best reference signal, improving recognition accuracy in challenging acoustic conditions.
2Measurement precision
If multiple speech recognition apparatuses are used to improve accuracy, then the system complexity increases
Solution Approach 1:
The patent merges the functionality of multiple speech recognition apparatuses into a coordinated system where each apparatus calculates its own SNR and participates in a selection process. The apparatuses work together to identify the reference signal, combining their computational resources to achieve higher accuracy without requiring all apparatuses to perform full recognition independently.
Solution Approach 2:
The SNR calculation acts as an intermediary mechanism that mediates between multiple apparatuses and the final recognition decision. Rather than requiring complex coordination protocols, the SNR serves as a simple intermediary metric that enables apparatuses to self-evaluate and select the best reference signal automatically.
3Productivity
If speech signals are processed locally on resource-constrained apparatuses, then processing capability is limited
Solution Approach 1:
Instead of requiring full speech recognition processing on all apparatuses, the system performs only the necessary SNR calculation locally and uses this partial action to determine the reference signal. This partial processing approach reduces energy consumption on battery-powered devices while still achieving accurate recognition by leveraging the computational resources of more capable apparatuses when needed.
Data Source
AI summary
A method and apparatus for speech recognition are provided. The method and the apparatus calculate signal to noise ratios (SNRs) of speech signals from a user received at speech recognition apparatuses. The method and the apparatus recognize a reference speech signal having a maximum SNR among the SNRs.