Sleep apnea is a dishevelment that causes interruption in breath or shoal of the respiration. The respiratory signal is classified into three states such as normal respiration, motion artifacts, and sleep apnea and it is obtained from a physionet. Firstly, using a second order auto regressive modeling, an algorithm is developed to attain the energy and frequency parameters of the signal and then the signal is classified with threshold based manual classification into any of the above taxonomy. In addition to this dataset, MLP is trained with a back propagation learning algorithm that results in reduced time, iterations and errors. Consequently, the training of SVM, a binary classifier used to solve multiple class problems is done with the same data set and classification is made to reduce overall errors. The overall efficiency of the above techniques is compared.
Keywords
Feature Extraction, Autoregressive Model, Burgs Method, Multilayer Perceptron, Back Propagation Learning Algorithm, Support Vector Machine
User
Information