Sparse Representation for Classification of Posture in Bed

Abstract

Redundant dictionaries, also known as frames, offera non–orthogonal representation of signals, which leads to sparsityin their representative coefficients. As this approach providesmany advantageous properties it has been used in various applicationssuch as denoising, robust transmissions, segmentation,quantum theory and others. This paper investigates the possibilityof using sparse representation in classification, comparing theachieved results to other commonly used classifiers. The differentmethods were evaluated in a real-world classification task inwhich the position of a lying patient has to be deduced basedon the data provided by a pressure mattress of 30×11 sensors.The investigated method outperformed most of the commonlyused classifiers with accuracy exceeding 92%, while being lessdemanding on design and implementation complexity.