Algorithm Classifying Depressed Patients’ Speech Samples Using Deep Learning

This research uses deep learning technique to develop a depression classifying algorithm from speech data of female subjects. The subjects in study were: Female subjects recovered from depression, female subjects with depression, and female subjects with high risk suicide. The objectives of research are to design and develop an algorithm to screen the depression patients from normal people through speech processing and deep learning, to find the classifying efficiency of depression patients to develop the algorithm, and to develop algorithms that can be used as medical assistive tool that can be used in medical practice in the future. We compared the performance of three Deep learning models: Convolutional Neural Networks (CNNs), Support Vector Machines (SVMs), and Recurrent Neural Networks with Long-Short Term Memory (RNN-LSTM). The results showed that the CNN model has performed the best classification achieving the high precision of 97%, recall of 92%, F1-score of 95%, and accuracy of 95%. This project shows that thedepression classifying algorithm using deep learning performs with highly satisfactory achievement and it could be applied in related medical research and beneficial to the field of medical innovation in the future.