Spectral Estimation Using Adaptive Sample Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for spectral estimation from unevenly spaced and irregularly sampled data, such as ARMA, DFT, and MF techniques, are inefficient for real-time applications due to high computation times and the need for excessive data sets, making them unsuitable for manufacturing environments.
Innovation Solution
A method that selects initial sets of samples from a data stream and successively combines them until a target signal-to-noise ratio is achieved, using a least-squares approach with QR factorization to estimate the spectrum efficiently, allowing for accurate spectral estimation with a near-minimum number of samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If ARMA methods are used for spectral estimation, then the method can handle unevenly spaced data, but the computation time becomes excessively high
Solution Approach 1:
The patent segments the spectral estimation process into two distinct stages: (1) selecting a subset of evenly spaced samples from the unevenly spaced data, and (2) applying standard DFT to this subset. This segmentation allows the method to handle unevenly spaced data while avoiding the computationally intensive ARMA modeling, thus resolving the contradiction between adaptability and computation time.
Solution Approach 2:
The patent extracts only the necessary information from the unevenly spaced data by selecting a subset of evenly spaced samples. This extraction approach retains the essential spectral characteristics while discarding redundant information, enabling efficient DFT processing without requiring full ARMA computation.
2Loss of time
If DFT methods are used for spectral estimation, then the computation is faster, but extremely large sets of input data are required with small inter-sample time spacing
Solution Approach 1:
The patent applies partial action by using only a subset of samples (specifically, evenly spaced samples) rather than requiring the entire dataset. This partial sampling approach maintains sufficient spectral estimation accuracy while dramatically reducing the quantity of data needed, thus resolving the contradiction between computation speed and data quantity requirements.
3Measurement precision
If MF methods are used for spectral estimation, then the method can provide spectrum estimates, but no practical means of determining sufficient samples exists resulting in excessively long measurement and computation times
Solution Approach 1:
The patent incorporates feedback through the use of timestamps to track and evaluate the quality of spectral estimates as samples are added. This feedback mechanism allows the system to determine when sufficient samples have been collected to achieve the desired measurement precision, thereby avoiding excessively long measurement and computation times while maintaining accuracy.
4Use of energy by moving object
If randomized sampling is used to reduce data rate, then power consumption decreases, but spectral estimation accuracy becomes difficult to achieve
Solution Approach 1:
The patent applies local quality by treating different portions of the data stream differently: it selects and processes only the evenly spaced samples (a local subset) for spectral estimation while ignoring the irregularly spaced samples. This approach maintains spectral estimation accuracy using a manageable subset of data, thereby enabling low-power operation through randomized sampling without sacrificing measurement precision.
Data Source
AI summary
A spectrum of a set of samples from a data stream of sampled data is estimated until a targeted signal to noise ratio is achieved.