Detection of sleep apnea with chaotic sound features

Snoring is one of the most important symptom of the Obstructive Sleep Apnea Syndrome (OSAS). When apnea is able to be diagnosed only using the snore sounds, recording and analysis of snore signals will be able to perform in home environment without the necessity of laboratory. Thus, diagnosing snore apnea by benefiting from snore signal has great importance. In this study, based on chaotic structure of the snore sounds, Largest Lyapunov Exponent (LLE) and mean value of divergence curves parameters are used as features for classification of snore sounds. OSAS/simple snoring situations are classified by means of a feed forward neural network. When the two features used as inputs of the neural network, total classifier performance rate was obtained as %96,58.