Χρήση του μικροφώνου των smartphone για την αξιολόγηση της συναισθηματικής κατάστασης : μια πιλοτική μελέτη
Year: 2024
Type: article
Abstract: Depression is one of the most common mental health conditions which affects millions people globally. Despite its commonality it often goes under-diagnosed due to stigmas around mental health and limitations of current diagnostic methods. The human voice is characterized as a biomarker and can be used for the diagnosis of various conditions. This makes it a powerful tool which can aid in the diagnosis of depression. In conjunction with recent advances in Artificial Intelligence in the field of human voice processing, these technologies can be leveraged to provide a powerful framework to diagnose and provide aid and relief to people suffering from depression. With the use of edge devices, such as mobile phones or home assistants, a person can be diagnosed easily and referred to a mental health professional to receive the appropriate treatment. In this thesis a new approach for depression diagnosis is presented. With the use of state-of-the-art voice processing algorithms, the models are able to find patterns from a persons' vocal features such as pitch, tone and speech which can indicate the likelihood of a person suffering from depression. This method combines the use of a spectral based model, namely VGGish, and a Transformer based model, Wav2vec 2.0. This multi-modal architecture is able to capture both high-level audio representations and rich temporal dependencies and linguistic information. In addition, this model is used in application, where voice recordings are analyzed to predict the likelihood of the presence of depression. This can serve as a help to people and mental health professionals as an accessible way to monitor their mental health. While the focus of this thesis is the development of the Artificial Intelligence model, this application showcases the real-world applicability of this research. The results indicate that voice analysis is a reliable tool that can be used for the detection of mental health problems in individuals, highlighting the need for accessible and early depression diagnosis. This thesis contributes to the field of AI-driven voice based diagnosis in the field of mental health and the integration of such technologies in peoples' everyday life.
Source: Νημερτής
Author: Δενάζη, Ειρήνη
Language: English
Topic: Emotion and Mood Recognition
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7162349865