Discrete wavelet transform and support vector machine applied to pathological voice signals identification
ABSTRACT An algorithm able to classify pathological and normal voice signals based on Daubechies discrete wavelet transform (DWT-db) and support vector machines (SVM) classifier is presented. DWT-db is used for time-frequency analysis giving quantitative evaluation of signal characteristics to identify pathologies in voice signals, particularly nodules in vocal folds, of subjects with different ages for both male and female. After using a linear prediction coefficients (LPC) filter, the signals mean square values of a particular scale from wavelet analysis are entries to a nonlinear least square support vector machine (LS-SVM) classifier, which leads to an adequate larynx pathology classifier which over 95% of classification accuracy.
[Show abstract] [Hide abstract]
ABSTRACT: This article presents a non-invasive speech processing method for the assessment and evaluation of voice hoarseness. A technique based on time-scale analysis of the voice signal is used to decompose the signal into a suitable number of high-frequency details and extract the high-frequency bands of the signal. A discriminating measure, which measures the roll-off in power in the high-frequency bands of the signal, with respect to the decomposition index, is developed. The measure reflects the presence and degree of severity of hoarseness in the analyzed voice signals. The discriminating measure is supported by frequency-domain and time-series analyses of the high-frequency bands of normal and hoarse voice signals to provide a visual aid to the clinician or therapist. A database of sustained long vowels of normal and hoarse voices is created and used to assess the presence and degree of severity of hoarseness. The results obtained by the proposed method are compared to results obtained by perturbation analysis.Biomedical Signal Processing and Control 10/2008; 3(4):283-290. DOI:10.1016/j.bspc.2008.06.002 · 1.53 Impact Factor