Prediction of Bipolar Disorder with Voice Analysis using Machine Learning Techniques

Abstract

The change in the speech is the responsive and well-founded measure of the depression and obsession of the bipolar disorder. This analysis mainly focuses on perceiving the voice attributes and phone calls data is collected as it acts as a main search-space maker for bipolar clutters. By combining the voice features with the phone call data based on their behavioral activities, self- monitoring data control and illness activities the accuracy would increase to effective states. The voice attributes and smartphones collect the activities of sample phone data and self-monitoring data. These activities are the root cause of the expansion of two symptoms: depression and obsession. These symptoms were introduced by a researcher who was rendered with smartphones. The phone call data were examined through a statistical random forest algorithm. The states were extracted from daily phone calls and are classified using voice attributes. These attributes are more determined and accurate to classify the maniac states. The main subject in comparing the voice attributes and self-observed data with the behavioral activities of phone call data is that these attributes increase the efficiency, vulnerability, and definiteness of classifying the affective states. The techniques used to detect the voice features are support vector machine (SVM) random forest. the proposed system will enhance the performance of the prediction of all the techniques. By comparing all these techniques by finding the accuracy of each technique we can know which technique predicts more accurately.