Simultaneous learning of speech feature and segment for classification of Parkinson disease

Speech feature learning is very important for the design of classification algorithm of Parkinson's disease (PD). Existing speech feature learning method for classification of PD just pays attention to the speech feature. This paper proposed a novel hybrid feature learning algorithm which puts the features of all the speech segments of each subject together, thereby obtaining new and high efficient features without feature transformation. Firstly, hybrid features was constructed by combining features and segments. Secondly, high efficient hybrid feature selection was conducted by various criteria. Thirdly, the selected hybrid features were applied for classification of PD. Besides, various evaluation criteria are introduced into feature selection in this manuscript. Experimental results show that this proposed algorithm can obtain new features (hybrid feature) with satisfactory classification accuracy. The selected features are very stable and meaningful.