Parallel Speech Recognition with Confidence-Based Task Abort
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition systems often suffer from reduced accuracy and increased processing time due to the use of single recognition systems or sequential processing, which can be inefficient and time-consuming, especially in mobile device applications where latency and accuracy are critical.
Innovation Solution
Implementing multiple speech recognition systems (SRS) that process audio signals in parallel and abort incomplete tasks if the generated recognition results meet a confidence threshold, allowing for improved latency and accuracy by selecting the most confident results first.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple speech recognition systems process the audio signal in parallel, then recognition accuracy is improved, but processing time increases due to waiting for the slowest system
Solution Approach 1:
The patent applies partial action by allowing speech recognition tasks to be aborted before completion when a confidence threshold is met. Instead of requiring all parallel SRS to finish processing, the system accepts results from systems that complete first and meet the confidence threshold, thereby reducing processing time while maintaining recognition accuracy through selective acceptance of partial results
2Loss of time
If a single speech recognition system is used, then processing time is reduced, but recognition accuracy is limited to the accuracy of that single system
Solution Approach 1:
The patent merges multiple speech recognition systems into a unified parallel processing architecture where multiple SRS simultaneously process the same audio signal. The results from these merged systems are then evaluated against a confidence threshold, combining the strengths of multiple systems to achieve higher recognition accuracy while maintaining efficient processing through selective result acceptance
3Reliability
If multiple speech recognition systems completely process the audio signal, then comprehensive results are obtained, but the time required is determined by the slowest system
Solution Approach 1:
The patent introduces dynamic control to the speech recognition process by implementing confidence threshold-based abort mechanisms. The system dynamically adjusts the number of active recognition tasks based on real-time confidence evaluations, aborting unnecessary processing when sufficient confidence is achieved. This dynamic approach maintains reliability through comprehensive initial processing while dramatically improving productivity by eliminating waiting time for slower systems
Data Source
AI summary
The subject matter of this specification can be embodied in, among other things, a method that includes receiving an audio signal and initiating speech recognition tasks by a plurality of speech recognition systems (SRS's). Each SRS is configured to generate a recognition result specifying possible speech included in the audio signal and a confidence value indicating a confidence in a correctness of the speech result. The method also includes completing a portion of the speech recognition tasks including generating one or more recognition results and one or more confidence values for the one or more recognition results, determining whether the one or more confidence values meets a confidence threshold, aborting a remaining portion of the speech recognition tasks for SRS's that have not generated a recognition result, and outputting a final recognition result based on at least one of the generated one or more speech results.