Exploring Kernels in SVM-Based Classification of Larynx Pathology from Human VoiceVisa övriga samt affilieringar
2010 (Engelska)Ingår i: Proceedings of the 5th International Conference on Electrical and Control Technologies ECT-2010, May 6-7, 2010, Kaunas, Lithuania, Kaunas: KUT , 2010, s. 67-72Konferensbidrag, Publicerat paper (Refereegranskat)
Abstract [en]
In this paper identification of laryngeal disorders using cepstral parameters of human voice is investigated. Mel-frequency cepstral coefficients (MFCC), extracted from audio recordings, are further approximated, using 3 strategies: sampling, averaging, and estimation. SVM and LS-SVM categorize pre-processed data into normal, nodular, and diffuse classes. Since it is a three-class problem, various combination schemes are explored. Constructed custom kernels outperformed a popular non-linear RBF kernel. Features, estimated with GMM, and SVM kernels, designed to exploit this information, is an interesting fusion of probabilistic and discriminative models for human voice-based classification of larynx pathology.
Ort, förlag, år, upplaga, sidor
Kaunas: KUT , 2010. s. 67-72
Nyckelord [en]
Laryngeal disorder, Pathological voice, Voice processing, Mel-frequency cepstral coefficients, Sequence kernel, Principal canonical correlation, Monte-Carlo sampling, Kullback-Leibler divergence, Earth mover’s distance, GMM, SVM
Nationell ämneskategori
Datavetenskap (datalogi)
Identifikatorer
URN: urn:nbn:se:hh:diva-5961Scopus ID: 2-s2.0-84941687065OAI: oai:DiVA.org:hh-5961DiVA, id: diva2:352842
Konferens
The 5th International Conference on Electrical and Control Technologies 6-7 May 2010, Kaunas, Lithuania
2010-09-232010-09-222018-09-05Bibliografiskt granskad