Obstructive sleep apnea detection with nonlinear analysis of speech

•Deformation of the vocal tract components in OSA patients changes the nonlinear behavior of voice.•Nonlinear analysis of speech signals has considerable potential in OSA detection.•The performance of nonlinear properties belong to consonants is more successful than those of vowels in OSA/healthy classification.•OSA is determined with a test accuracy of 82.5% by using a few nonlinear features of consonant voices.