Vocal Fold Disorder Detection by applying LBP Operator on Dysphonic Speech Signal Ghulam Muhammad 1 , Zulfiqar Ali 1,3 , Mansour Alsulaiman 1 , Khalid Almutib 2 1 Department of Computer Engineering, 2 Department of Software Engineering King Saud University P.O. Box 51178, Riyadh 11543 SAUDI ARABIA {ghulam, zuali, msuliman, muteb}@ksu.edu.sa 3 Department of Electrical and Electronic Engineering Universiti Teknologi PETRONAS Bandar Seri Iskandar, 31750, Tronoh, Perak MALAYSIA Abstract: - The automatic system for voice pathology assessment is one of the active areas for researchers in the recent years due to its benefits to the clinicians and presence of a significant number of dysphonic patients around the globe. In this paper, a voice disorder detection system is developed to differentiate between a normal and pathological voice signal. The system is implemented by applying the local binary pattern (LBP) operator on Mel-weighted spectrum of a signal. The LBP is considered as one of the sophisticated techniques for the image processing. The technique also provided very good results for voice pathology detection during this study. The English voice disorder database MEEI is used to evaluate the performance of the developed system. The results of the LBP operator based system are compared with MFCC and found to be better than MFCC. Key-Words: - LBP operator, MFCC, Vocal fold disorders, Sustained vowel, MEEI database, disorder detection system. 1 Introduction Voice pathologies affect the vocal folds, and these disorders produce irregular vibrations in the vocal folds due to the malfunctioning of the voice box. Vocal fold pathologies exhibit variations in a vibratory cycle of the vocal folds due to their incomplete closure. The voice disorder also changes the shape of the vocal tract and produces irregularities in spectral properties. In the last decade, much research has been done on the automatic detection and classification of vocal fold diseases, and these tasks continue to require further investigation due to the lack of standard diagnosing approaches/equipment for voice disorders. Automatic voice pathology detection can be accomplished by various types of features, which can be obtained by the long-term and short-term signal analysis. The long-term parameter can be derived by acoustic analysis [1, 2] of speech. The short-term parameters are further divided into two groups: parametric features and non-parametric features [3]. The parametric features represent the resonant structure of the human vocal cord and can be obtained by linear prediction coefficients (LPC) [4], and LPC-based cepstrum (LPCC) [5]. The non- parametric features mimic the behavior of the human auditory system and can be derived from the FFT based Mel-frequency cepstral analysis (MFCC) [6, 7]. Various classification techniques of pattern recognition such as HMM (Hidden Markov Model) [8, 9], GMM (Gaussian Mixture Model) [10], Vector Quantization (VQ) [11], Support Vector Machine (SVM) [12], Multilayer Perceptron (MLP) [13], Neural Networks (NN) [14], k-means Nearest Neighbors (KNNs) [15], Linear Discriminant Analysis (LDA) [16], and Learning VQ (LVQ) [17] are used to detect and\or classify voice disorders. The vocal tract properties can be modeled using the all-pole model with the help of LPC features. These features represent the main vocal tract resonance properties in the acoustic spectrum. LPC highlights these formant structures for a speaker to Recent Advances in Intelligent Control, Modelling and Simulation ISBN: 978-960-474-365-0 222