Pathological Voice Classification Using Empirical Mode Decomposition and Ensemble Empirical Mode Decomposition with Extreme Learning Machine
Year: 2026
Type: article
Abstract: Pathological voice analysis has been proved as an important computer aided diagnostic approach for early identification of laryngeal disorders using sustained phonation recordings. Due to the highly non-linear and non-stationary nature of speech signals, traditional spectral features often fail to capture subtle changes related to pathological vocal fold vibration. In this paper, we propose a hybrid feature extraction framework by combining Empirical Mode Decomposition (EMD) and Ensemble Empirical Mode Decomposition (EEMD) to extract discriminative time-frequency representations from sustained vowel (/a/) recordings of the VOICED database. The decomposed intrinsic mode functions are further characterized by instantaneous energy and instantaneous frequency obtained from Hilbert spectral analysis and statistical descriptors that preserve temporal and spectral characteristics of healthy and pathological voices. The resulting feature vectors are classified using an Extreme Learning Machine (ELM) due to its fast learning, good generalization performance and low computational complexity. The experimental evaluation is carried out using standard performance measures such as accuracy, sensitivity, specificity, precision and F1-score. The hybrid EMD–EEMD feature set can provide complementary information to the individual decomposition methods and can mitigate mode-mixing effects, which demonstrates better performance than each individual decomposition method. The proposed EMD+EEMD-ELM framework obtained the overall classification accuracy of 98.74%, sensitivity, specificity, precision and F1-score of 98.42%, 99.05%, 98.63% and 98.52% respectively on the VOICED dataset under the adopted experimental protocol.
Author: Anil Kumar V
Institution: REVA University
Language: English
Topic: Voice and Speech Disorders
Subfield: Speech and Hearing
Field: Health Professions
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7202035549