Spectral Classification of Oral and Nasal Snoring Sounds Using a Support Vector Machine
Autor: | Kazuya Yonezawa, Masashi Furukawa, Yohichiro Kojima, Masahito Yamamoto, Tsuyoshi Mikami |
---|---|
Rok vydání: | 2013 |
Předmět: |
Human-Computer Interaction
Support vector machine Artificial Intelligence business.industry Computer science Pattern recognition Computer Vision and Pattern Recognition Artificial intelligence business Stellar classification psychological phenomena and processes nervous system diseases respiratory tract diseases |
Zdroj: | Journal of Advanced Computational Intelligence and Intelligent Informatics. 17:611-621 |
ISSN: | 1883-8014 1343-0130 |
Popis: | Since oral breathing during sleep tends to make the upper airway more collapsible, loud snoring caused by oral breathing is found in many sleep apnea/hypopnea patients and should be detected in the earlier stage. But unfortunately we cannot know our own sleep condition or snoring. Thus, a simple method that can detect oral snoring makes it possible to become a useful technique to develop a home medical device. For such purpose, we adopt a Support Vector Machine (SVM) classifier so as to classify oral and nasal snoring sounds based on the spectral properties. Fifteen subjects are asked to simulate snoring with oral and nasal breath respectively and the sounds are recorded with a linear sound recorder. We adopted seven kernel functions (linear, polynomial, sigmoid, Gaussian, Laplacian, chisquare, and Kullback-Leibler) for SVM-based spectral classification. As a result, over 95% of snoring sounds are successfully classified under the various cross validation test. |
Databáze: | OpenAIRE |
Externí odkaz: |