Sparse signal reconstruction using gradient-threshold based method

The performance of gradient (steepest descent) and the threshold-based algorithms are observed in terms of the sparse signal reconstruction. The advantages of both methods are combined within the new approach used to recover all samples from randomly under-sampled signal. The gradient-based algorith...

Full description

Saved in:
Bibliographic Details
Published in2018 7th Mediterranean Conference on Embedded Computing (MECO) pp. 1 - 4
Main Authors Zaric, Maja Lakicevic, Draganic, Andjela, Orovic, Irena, Stankovic, Srdjan
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.06.2018
Subjects
compressive sensing
Convex functions
Discrete Fourier transforms
Embedded computing
Frequency measurement
gradient based algorithm
Optimization
Signal reconstruction
Spectral analysis
threshold
variance
Online AccessGet full text

Cover

More Information
Summary:The performance of gradient (steepest descent) and the threshold-based algorithms are observed in terms of the sparse signal reconstruction. The advantages of both methods are combined within the new approach used to recover all samples from randomly under-sampled signal. The gradient-based algorithm may fail to recover the signal unless a relatively large number of iterations is performed, which can be time consuming. The procedure can be speed up by stopping the gradient algorithm at certain convenient iteration and continuing with the reconstruction using the threshold-based method. Threshold is calculated in a way to separate the signal components from the spectral noise that is still left in the signal after the gradient-based reconstruction. The exact values of the signal amplitudes are calculated by solving the optimization problem. The proposed method increases the reconstruction speed with satisfactory reconstruction accuracy. The theory is proved with experiments.
DOI:10.1109/MECO.2018.8406090