Adaptive Signal Acquisition With Distilled Sensing for Weak Sparse Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing signal recovery methods struggle to reliably identify sparse signals in noise, particularly when the signal-to-noise ratio is low, as they fail to control both false-discovery and non-discovery proportions effectively, especially in adaptive measurement scenarios where signals are weak or have few relevant components.
Innovation Solution
The method employs a multi-part data acquisition procedure called Distilled Sensing, which involves iterative sensing with increasing observation time and energy allocation, focusing on likely signal components and ignoring noise, to improve signal detection and estimation by systematically refining measurements based on previous outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If coordinate-wise thresholding is used for signal recovery, then the method is simple and computationally efficient, but it cannot reliably control both false-discovery proportion and non-discovery proportion when signals are weak
Solution Approach 1:
The patent segments the signal recovery process into multiple sequential stages, each with its own thresholding operation. Instead of a single thresholding step, the method performs iterative thresholding where each stage refines the estimate from the previous stage, allowing progressive improvement in reliability while maintaining computational feasibility through modular processing
Solution Approach 2:
The patent employs dynamic thresholding where the threshold level adapts across multiple stages of processing. The threshold is not fixed but evolves through iterative applications, with each stage using information from previous stages to adjust subsequent thresholding parameters, enabling reliable control of both false-discovery and non-discovery proportions
2Reliability
If observation time and energy are increased to improve signal detection, then detection reliability improves, but acquisition time and energy expenditure increase
Solution Approach 1:
The patent performs preliminary coarse thresholding to identify candidate signal components before investing more observation time and energy. By first making a rough identification pass and then focusing resources only on promising candidates in subsequent stages, the method achieves high detection reliability without requiring uniformly increased observation time across all components
Solution Approach 2:
The patent applies observation time and energy selectively rather than uniformly. It concentrates measurement resources on components identified as likely signal carriers in previous stages, applying excessive observation effort only where needed to confirm signal presence, while using minimal or no additional resources on components already confidently identified or ruled out
Data Source
AI summary
Methods for adaptive data acquisition are disclosed herein. In one aspect, methods for adaptive data acquisition include performing a first sensing method on a signal having a plurality of components to determine the likelihood that a component is not a relevant signal component, retaining a portion of the signal components sensed using the first sensing method that are above a first threshold, performing a second sensing method on the signal components retained above a first threshold to determine the likelihood that a component is not a relevant signal component, wherein the second sensing method is more reliable than the first sensing method at determining the likelihood that a component is not a relevant signal component, and retaining a portion of the signal components sensed using the second sensing method that are above a second threshold.