Abstract
This paper proposes the use of a new feature extraction technique named I-Vectors to detect and classify Dysphonic voice data. After initial pre-processing and framing, a total of 52 important acoustic features are extracted from the speech. Gaussian Mixture Model (GMM) algorithm makes use of an Expectation–Maximization (EM) algorithm to compute the zeroth, first and second order statistics pertaining to the UBM model for final extraction of I-Vectors. This high dimensional supervector is transformed into a low dimensional subspace called the Total Variability space (TV subspace) following which a new vector set (called I-Vectors) in extracted using the UBM model and Baum-Welch statistics. A total of 110 I-Vectors are extracted for the given input data after undergoing Cholesky whitening. This newly created data is now used for classifying normal and dysphonia samples. The application of Support Vector Machines (SVM) along with the extracted I-Vectors gives 98% accuracy on the unknown data set, which is a significant improvement in accuracy, as compared to 92% obtained without using I-Vectors.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
