Statistical Compressive Sensing for Efficient Signal Reconstruction and Classification

Abstract

Compressive sensing (CS) represents a signal processing technique for simultaneous signal acquisition and compression that relies on signal dimensionality reduction. Statistical compressive sensing (SCS) uses statistical models to develop an efficient sampling strategy for signals that follow some statistical distribution. In this paper, statistical model based on Gaussian mixtures is employed to design an efficient framework for the CS signal reconstruction and classification. A robust classification method based on sparse signal representation using overcomplete eigenvector dictionaries andl <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -norm is presented. Optimal non-adaptive measurement matrix for observed Gaussian mixture model is discussed. A series of experiments to analyze the performance of the proposed method has been performed and presented in the experimental results section.