Compressed Signal Reconstruction Using Past-Signal Dictionary Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing compressed sensing technologies face a trade-off between reconstruction accuracy and speed, with methods like BSBL achieving high accuracy but at the cost of increased reconstruction time, while algorithms like OMP provide fast reconstruction but with lower accuracy.
Innovation Solution
A signal reconstruction apparatus that uses a dictionary matrix populated with past signals of the same type as the target signal, allowing for highly accurate and high-speed reconstruction by converting the target signal into a sparse vector.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If BSBL reconstruction algorithm is used, then reconstruction accuracy is improved, but reconstruction time increases
Solution Approach 1:
The patent applies preliminary action by pre-processing the observation matrix to identify and retain only significant elements above a threshold value, creating a sparse observation matrix in advance. This pre-filtering reduces the computational burden during the actual BSBL reconstruction process, allowing high accuracy to be maintained while significantly reducing reconstruction time.
Solution Approach 2:
The patent extracts and removes insignificant elements from the observation matrix by applying a threshold criterion. Elements below the threshold are set to zero, creating a sparse matrix that contains only the most important information. This extraction process eliminates redundant computations while preserving the essential signal characteristics needed for accurate reconstruction.
2Measurement precision
If dictionary matrix with more columns is used, then reconstruction accuracy is improved, but reconstruction time increases
Solution Approach 1:
The patent extracts only the significant columns from the dictionary matrix by applying a threshold to the observation matrix. This selective extraction creates a sparse dictionary that contains only the most relevant basis vectors, reducing the number of columns needed to achieve accurate reconstruction while decreasing computational time.
Solution Approach 2:
The patent changes the parameter of the dictionary matrix by dynamically adjusting the number of columns based on the sparsity level and significance threshold. Instead of using a fixed large dictionary, the system adapts the dictionary size to match the actual complexity of the signal, optimizing both accuracy and speed.
3Productivity
If observation matrix with more elements is used, then compression capability is improved, but power consumption increases
Solution Approach 1:
The patent extracts and removes redundant elements from the observation matrix by applying a threshold criterion. This creates a sparse observation matrix that maintains compression capability while reducing the number of non-zero elements that require computational processing, thereby lowering power consumption in resource-constrained sensing devices.
Solution Approach 2:
The patent changes the density parameter of the observation matrix by controlling the threshold level. By adjusting this parameter, the system can balance between compression ratio and computational complexity, enabling effective compression with reduced power consumption by optimizing the sparsity level of the matrix.
Data Source
AI summary
A measurement signal processing unit includes a reception part that receives a compression signal y obtained by using an observation matrix Φ from a time-series detection signal of a measurement target, and a reconstruction part that reconstructs the received compression signal y. The reconstruction part includes the observation matrix Φ and a dictionary matrix Ψ including a past signal of the measurement target, obtains an estimation vector {circumflex over ( )}s by inputting the compression signal y and a sensing matrix θ to a reconstruction algorithm execution module, and derives a reconstruction signal {circumflex over ( )}x corresponding to a detection signal x by inputting, to a calculation module for obtaining a product, the obtained estimation vector {circumflex over ( )}s and the dictionary matrix Ψ.