Depression and Anxiety Screening via Free Conversational Speech in Naturalistic Condition

The Depression and Anxiety Screening via Free Conversational Speech in Naturalistic Conditions project aims to develop an innovative solution for the early detection of mental health issues such as depression and anxiety. By leveraging speech analysis in a naturalistic setting, the system evaluates conversational speech to identify signs of mental distress, offering a non-invasive and accessible method for mental health screening. The project involves audio pre-processing techniques like normalization and resampling to prepare .wav files for analysis. Key features such as Mel-Frequency Cepstral Coefficients (MFCC) and spectrograms are extracted from the speech, capturing patterns related to emotional tone, speech cadence, and rhythm, which are linked to emotional well-being. Machine learning models like Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and Light Gradient Boosting Machine (LightGBM) are trained separately to classify the speech as depression/anxiety or normal mental health, outputting a simple Yes/No prediction. The system's front-end is developed using HTML, CSS, and JavaScript, providing an easy-to-use interface for users to record and submit their speech. The back-end is powered by Python and Flask, processing audio, training models, and delivering results. This system is particularly useful for remote mental health screening, offering an efficient and accessible way to monitor emotional well-being. Early detection of mental health issues during pregnancy could improve maternal health, enable timely interventions, and contribute to the well-being of both the mother and the baby.

Keywords:

Depression, Anxiety, Conversational Speech, Speech Analysis, Machine Learning, Mel-Frequency Cepstral Coefficients (MFCC), Feature Extraction, Sentiment Classification, Flask Framework