Decoder Output Combining for Recognition Accuracy and Power
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data pattern recognition systems, such as speech recognition, face increased power consumption and inaccurate results when operating multiple recognition engines in parallel, leading to degraded battery life and suboptimal solutions.
Innovation Solution
A method to combine results from multiple decoders by generating a network of paths based on their outputs, scoring them to find the highest path score, calculating weighting factors for each decoder, and re-scoring to refine the path score, thereby generating an accurate decoded solution while minimizing power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple recognition engines are operated in parallel to improve accuracy, then recognition accuracy is improved, but power consumption increases proportionately
Solution Approach 1:
The system performs preliminary actions by having multiple decoders generate outputs and build a network of paths in advance. The combining module then scores these pre-generated paths to identify the highest scoring path, avoiding the need to re-run decoding processes and reducing power consumption while maintaining accuracy improvements.
Solution Approach 2:
The combining module acts as an intermediary that receives outputs from multiple decoders, scores the generated paths, and selects the best path. This intermediary component allows the system to leverage multiple decoders for accuracy while avoiding the full power consumption of running multiple complete recognition engines.
2Measurement precision
If multiple recognition engines are operated in parallel to improve accuracy, then recognition accuracy is improved, but device complexity increases
Solution Approach 1:
The system segments the recognition process into distinct components: multiple decoders that generate outputs and the combining module that scores and selects paths. This segmentation allows each component to perform a specific function, improving accuracy through multiple decoders while managing complexity by clearly defining the role of each segment.
Solution Approach 2:
The combining module merges the outputs from multiple decoders by scoring paths that combine these outputs. This merging approach allows the system to leverage the strengths of multiple decoders while using a unified scoring mechanism to select the best combined path, thereby improving accuracy without proportionally increasing system complexity.
3Measurement precision
If multiple recognition engines are operated in parallel, then solution accuracy is improved, but battery lifetime is degraded
Solution Approach 1:
The system performs preliminary decoding actions with multiple decoders and pre-generates the network of paths. By scoring these pre-generated paths without re-running the decoding processes, the system reduces energy consumption and extends battery lifetime while still achieving improved solution accuracy through the combination of multiple decoder outputs.
Solution Approach 2:
The system discards the need to re-run complete decoding processes by recovering and reusing the outputs already generated by multiple decoders. The combining module scores these existing outputs to find the best path, thereby avoiding redundant energy-consuming operations and extending battery lifetime while maintaining accuracy improvements.
Data Source
AI summary
Embodiments include a method, apparatus, and a computer program product for combining results from multiple decoders. For example, the method can include generating a network of paths based on one or more outputs associated with each of the multiple decoders. The network of paths can be scored to find an initial path with the highest path score based on scores associated with the one or more outputs. A weighting factor can be calculated for each of the multiple decoders based on a number of outputs from each of the multiple decoders included in the initial path with the highest path score. Further, the network of paths can be re-scored to find a new path with the highest path score based on the scores associated with the one or more outputs and the weighting factor for each of the multiple decoders.