Hierarchical parallel PSO-SVM based subject-independent sleep apnea classification

Abstract

This paper presents a method for subject independent classification of sleep apnea by a parallel PSO-SVM algorithm. In the proposed structure, swarms are separated into masters and slaves and accessing to the global information is restricted according to their types. Biosignal records that used as the input of the system are air flow, thoracic and abdominal respiratory movement signals. The classification method consists of the three main parts; feature generation, feature selection and data reduction based on parallel PSO-SVM, and the final classification. Statistical analyses on the achieved results show efficiency of the proposed system. © 2012 Springer-Verlag.