Comparison of ANFIS and M-SVMs Classifiers in Silence-Breathing-Snore Sounds Classification

ANKIŞHAN H., ARI F.

9th International Conference on Electronics Computer and Computation (ICECCO 2012), Ankara, Türkiye, 1 - 03 Kasım 2012, ss.207-210 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Ankara
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.207-210
  • Anahtar Kelimeler: ANFIS, multi class support vector machines, snore related sound classification
  • Ankara Üniversitesi Adresli: Evet

Özet

Snoring is a disease which causes collapse in upper airway while sleeping. There have been many studies to diagnosis and classify these problem in the literature by using snore related sounds (SRSs). Some have only classified SRSs into snore/non-snore, while others have diagnosed apnea/hypopnea disorders by using different classifiers. The purpose of this study is to classify SRSs into simple snore, breathing (respiratory sounds), and silent episodes by using adaptive neuro fuzzy inference system (ANFIS) and multi class support vector machines (M-SVMs). In the experimental study, SRSs are segmented as snore, breathing (respiratory sounds), and silent parts. After segmentation, features are calculated for each segment part and used as classifiers' input. It is shown with experimental results that non-linear separable algorithms of M-SVMs and ANFIS were able to classify the SRSs. Accuracy of testing rates for M-SVMs and ANFIS are estimated at 93.94% and 92.28, respectively.