Idle Resource Speech Recognition via Dynamic Algorithm Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition systems are inefficient as they often underutilize hardware during off-peak periods, leading to imperfect performance and additional costs due to the need for hardware to handle peak load conditions, which are rare.
Innovation Solution
Implementing a system that runs multiple recognizers in parallel with different characteristics during off-peak times to improve recognition accuracy by combining their outputs, utilizing idle resources without additional hardware or cost, and dynamically scaling complexity based on demand.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If hardware is provisioned to handle peak load demands, then system reliability is improved, but hardware utilization efficiency deteriorates
Solution Approach 1:
The system dynamically adjusts the complexity and computational resources allocated to speech recognition processing based on real-time demand conditions. During peak load periods, the system uses standard recognition algorithms with limited resources, while during off-peak periods, it activates more complex and accurate recognition algorithms that utilize otherwise idle hardware resources, thereby improving both reliability and utilization efficiency
2Measurement precision
If more complex recognition algorithms are used, then recognition accuracy is improved, but computational load increases
Solution Approach 1:
The system implements periodic variation in algorithmic complexity based on temporal patterns of demand. During off-peak periods when computational load is low, the system periodically activates more complex and accurate recognition algorithms. During peak periods when computational load is high, the system switches to simpler algorithms, thereby achieving high accuracy when conditions permit while maintaining system responsiveness
3Measurement precision
If hardware resources are increased to improve recognition performance, then recognition accuracy is improved, but system cost increases
Solution Approach 1:
The system enables idle hardware resources to serve dual purposes: handling routine speech recognition tasks during peak periods and providing additional computational capacity for enhanced recognition algorithms during off-peak periods. This self-service approach allows the existing hardware infrastructure to improve recognition accuracy without requiring additional hardware investments or increased system cost
Data Source
AI summary
Disclosed herein are systems, methods, and computer-readable storage media for improving automatic speech recognition performance. A system practicing the method identifies idle speech recognition resources and establishes a supplemental speech recognizer on the idle resources based on overall speech recognition demand. The supplemental speech recognizer can differ from a main speech recognizer, and, along with the main speech recognizer, can be associated with a particular speaker. The system performs speech recognition on speech received from the particular speaker in parallel with the main speech recognizer and the supplemental speech recognizer and combines results from the main and supplemental speech recognizer. The system recognizes the received speech based on the combined results. The system can use beam adjustment in place of or in combination with a supplemental speech recognizer. A scheduling algorithm can tailor a particular combination of speech recognition resources and release the supplemental speech recognizer based on increased demand.