Automatic detection of neurological disordered voices using mel cepstral coefficients and neural networks

Acoustical voice analyses and measurement methods might provide useful biomarkers for the diagnosis of neurological disordered voices. This paper presents a method for automatic detection of neurological disordered voices like Parkinson's disease, cerebellar demyelination and stroke using the Mel-frequency cepstral coefficient (MFCC) features. The features extracted were given to a multilayer neural network and trained to classify whether the voice was neurological disordered or normal subject. There are no risks involved in capturing and analysis of voice signals as it is noninvasive by nature and in carefully controlled circumstances, it can provide a large amount of meaningful data. The data collected in the present work consist of 137 sustained vowel phonations (/ah/), among them 73 phonations are from patients suffering from different neurological diseases and 64 phonations from controlled subjects including both male and female subjects. Thirteen MFCC features are used as input to the optimally designed artificial neural network (ANN) for classification. 112 phonations were used to train the network and 25 phonations for testing. The best classification accuracy achieved was 92%.