The characteristics of speech and voice in neurological diseases, such as, Parkinson's disease (PD), cerebellar demyelination, senile disease and stroke, have a realistic potential to provide information for early detection of onset, progression, and severity of these diseases. 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 136 sustained vowel phonations (/ah/), among them 83 phonations are from patients suffering from different neurological diseases and 53 phonations from controlled subjects including both male and female subjects. A total of 16 features were extracted from the voice data and significant differences between the two group ‘means’ were evaluated using student's t-test. Significant findings in measurements were found in all types of shimmers and jitters features, except in measures of pitch. Further, all the 16 features were used as input to the artificial neural network (ANN) for classification. Two types of ANN are used for classification, the multilayer perceptron (MLP) network and radial basis function (RBF) network. 112 phonations were used to train the network and 24 phonations for testing. The RBF network gave a better classification with 90.12% for training set and 87.5% for test set compared to MLP with 86.66% for training set and 83.33% for test set.
展开▼
