Pathological Voice Classification Based on a Single Vowel's Acoustic Features

This paper mainly studies on the classification of pathological voice from normal voice based on the sustained vowel /a/. Firstly, the original 18 acoustic features are extracted. Then on the basis of the extracted parameters, this paper recognizes the pathological voice using AD Tree. During the classification stage, the cross-validation of features is also as references in the process. This method is validated with a sound database provided by the Massachusetts Eye and Ear Infirmary (MEEI). After the 10 fold cross-validation, comparing with 7 other kinds of classifiers, the experimental results show that AD Tree can get the highest recognition rate of 95.2%. The method in this paper shows that all the extracted parameters are reasonable in the following recognition process and AD tree is a good recognition way in pathological voice research.

Download Full-text

Open Quotient Influence on Performances of Pathological Voice Classification by SVM

Second International Conference on Advances in Information Processing and Communication Technology - IPCT 2015 ◽

10.15224/978-1-63248-044-6-120 ◽

2015 ◽

Author(s):

AICHA BOUZID ◽

ASMA BELHAJ ◽

NOUREDDINE ELLOUZE

Keyword(s):

Voice Classification ◽

Pathological Voice

Download Full-text

Pathological Voice Classification Using Mel-Cepstrum Vectors and Support Vector Machine

2018 IEEE International Conference on Big Data (Big Data) ◽

10.1109/bigdata.2018.8622208 ◽

2018 ◽

Cited By ~ 1

Author(s):

Maryam Pishgar ◽

Fazle Karim ◽

Somshubra Majumdar ◽

Houshang Darabi

Keyword(s):

Support Vector Machine ◽

Support Vector ◽

Voice Classification ◽

Pathological Voice ◽

Mel Cepstrum

Download Full-text

Non-intrusive technique for pathological voice classification using jitter and shimmer

2017 International Conference on Computer, Communication and Signal Processing (ICCCSP) ◽

10.1109/icccsp.2017.7944104 ◽

2017 ◽

Cited By ~ 2

Author(s):

N Sripriya ◽

S Poornima ◽

R Shivaranjani ◽

Preethi Thangaraju

Keyword(s):

Voice Classification ◽

Pathological Voice

Download Full-text

Voice Signal Analysis with the Application in Biomedicine

Sensor Letters ◽

10.1166/sl.2020.4187 ◽

2020 ◽

Vol 18 (2) ◽

pp. 122-127

Author(s):

Vikas Mittal ◽

R. K. Sharma

Keyword(s):

Signal Analysis ◽

Voice Quality ◽

Support Vector ◽

Efficiency Improvement ◽

Average Efficiency ◽

Voice Signal ◽

Signal Parameters ◽

Nearest Neighbours ◽

Voice Classification ◽

Pathological Voice

Voice pathology is the result of improper vocal use. Poor vocal exercise and repeated laryngeal infection may lead to worse voice quality and vocal stresses. This work uses glottal signal parameters obtained from speakers of distinct ages to identify voice disorders. The parameters obtained from the glottal signal, Mel Frequency Cepstrum Coefficients (MFCCs) and combination of glottal and MFFCs are used for pathological voice classification. Support Vector Machine (SVM) and K-Nearest Neighbours (KNN) algorithms are used. Results show that best classification results are achieved using combinations of MFFCs and with glottal parameters including MOQ, which is a novel outcome and most important involvement of this study, with an average efficiency improvement of 3%.

Download Full-text