Pathological Voice Detection Using Efficient Combination of Heterogeneous Features

Ji-Yeoun LEE  Sangbae JEONG  Minsoo HAHN  

Publication
IEICE TRANSACTIONS on Information and Systems   Vol.E91-D   No.2   pp.367-370
Publication Date: 2008/02/01
Online ISSN: 1745-1361
DOI: 10.1093/ietisy/e91-d.2.367
Print ISSN: 0916-8532
Type of Manuscript: LETTER
Category: Speech and Hearing
Keyword: 
pathological voice detection,  heterogeneous feature combination,  mel-frequency filter bank energies,  higher-order statistics,  pattern classification algorithm,  

Full Text: PDF(365.6KB)>>
Buy this Article




Summary: 
Combination of mutually complementary features is necessary to cope with various changes in pattern classification between normal and pathological voices. This paper proposes a method to improve pathological/normal voice classification performance by combining heterogeneous features. Different combinations of auditory-based and higher-order features are investigated. Their performances are measured by Gaussian mixture models (GMMs), linear discriminant analysis (LDA), and a classification and regression tree (CART) method. The proposed classification method by using the CART analysis is shown to be an effective method for pathological voice detection, with a 92.7% classification performance rate. This is a noticeable improvement of 54.32% compared to the MFCC-based GMM algorithm in terms of error reduction.