Sparse S-Transform Signal Processing for Memory Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The standard Fourier analysis and existing Time-Frequency Representations, including the S-transform, require substantial computer memory and longer execution times, making them impractical for clinical and many industrial applications due to high system requirements.
Innovation Solution
A sparse approximation of the S-transform is implemented, which reduces memory storage and processing time by retaining fewer pixels at low frequencies, decimating voices at high frequencies, and using interpolation to reconstruct signal data, allowing for efficient signal processing in the Stockwell domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the full S-transform is used for signal processing, then time-frequency localization accuracy is improved, but memory storage requirements and processing time increase substantially
Solution Approach 1:
The patent extracts only the essential components of the S-transform by retaining fewer pixels at low frequencies and decimating voices at high frequencies. This selective extraction maintains the critical time-frequency localization information while substantially reducing the data volume stored in memory, directly resolving the contradiction between accuracy and storage requirements.
Solution Approach 2:
The patent applies partial action by using a sparse approximation that processes only the most significant frequency components rather than the complete S-transform. This partial processing approach maintains sufficient localization accuracy for practical applications while dramatically reducing memory and computational resources required.
2Measurement precision
If the full S-transform is used for signal processing, then time-frequency localization accuracy is improved, but execution time increases substantially
Solution Approach 1:
The patent extracts only the essential components of the S-transform by retaining fewer pixels at low frequencies and decimating voices at high frequencies. This selective extraction maintains the critical time-frequency localization information while substantially reducing the data volume stored in memory, directly resolving the contradiction between accuracy and storage requirements.
Solution Approach 2:
The patent applies partial action by using a sparse approximation that processes only the most significant frequency components rather than the complete S-transform. This partial processing approach maintains sufficient localization accuracy for practical applications while dramatically reducing memory and computational resources required.
3Productivity
If standard Fourier analysis is used, then computational efficiency is maintained, but time-frequency localization capability is insufficient
Solution Approach 1:
The patent segments the frequency spectrum into different ranges (low frequencies and high frequencies) and applies different processing strategies to each segment. This segmentation allows the system to maintain computational efficiency by using simpler processing for most frequencies while applying more sophisticated time-frequency analysis only where necessary, thus achieving both efficiency and localization capability.
Solution Approach 2:
The patent applies local quality by providing different levels of time-frequency analysis at different frequency ranges. At low frequencies, more detailed analysis is applied where time-frequency localization is most critical, while high-frequency components are processed more coarsely. This localized approach maintains computational efficiency while providing sufficient time-frequency localization capability where needed.
Data Source
AI summary
The present invention relates to a method and system for processing time series signal data or image signal data indicative of a characteristic of an object. Received signal data are transformed into second signal data within a Stockwell domain based upon a sparse approximation of a S-transform of the signal data. The second signal data are then processed within the Stockwell domain to extract features therefrom. The processing includes determination of local spectra at predetermined locations; determination of voices at predetermined frequencies; and filtering of the signal data using a filter function in dependence upon frequency and time or space. The signal processing method and system according to the invention enables signal processing based on the S-transform using a desktop computer or workstation.