A Data Reconstruction Algorithm Based on Neural Network for Compressed Sensing

Citations Over Time

Abstract

A multi-layers neural network is built to reconstruct the original data from the compressed sensed data in wireless sensor networks. Unlike the classical data reconstruction algorithms in compressed sensing theory, such as convex optimization based algorithms and greedy algorithms, the proposed data reconstruction algorithm is inspired by the unsupervised learning algorithm in machine learning. The proposed architecture of the Artificial Neural Network (ANN) is trained by minimizing the errors between input data and output data. The experiments were carried out based on real-world sensed dataset and the results demonstrate that the proposed algorithm presents a higher data reconstruction accuracy than the OMP and IHT algorithms. Meanwhile, the data reconstruction speed of the proposed algorithm is also faster than its counterparts.

Related Papers

• → Compressive Sensing for streaming signals using the Streaming Greedy Pursuit(2010)15 cited
• → Image Reconstruction from Patch Compressive Sensing Measurements(2018)7 cited
• Study on Performance of Greedy Algorithms in Compressive Sensing(2014)
• Study on Performance of Greedy Algorithms for Signal Reconstruction(2013)