International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 10 Issue: 07 | July 2023
p-ISSN: 2395-0072
www.irjet.net
Mental Illness Prediction based on Speech using Supervised Learning Nandini B M1, Ankita Vishwanatha Hegde2, Bhumika Shetty2, N Aditya Bhat 2, Rakesh R2 1Assistant Professor, Dept. of Information Science & Engineering, National Institute of Engineering, Mysuru,
Karnataka, India
2Student, Dept. of Information Science & Engineering, National Institute of Engineering, Mysuru, Karnataka, India
---------------------------------------------------------------------***--------------------------------------------------------------------Abstract - This research presents a novel approach to accurately detect depression using voice-based analysis and machine
learning algorithms. The main objective is to create a model that predicts whether an individual is experiencing depression using their audio inputs. The platform allows users to provide voice recordings, which are analysed using various acoustic features. By employing a supervised learning algorithm, the model effectively classifies the recordings as indicative of mental illness or not, thereby achieving high accuracy in predicting depression from voice recordings. The findings of this research demonstrate the potential of voice-based analysis and machine learning algorithms in early diagnosis and treatment of depression. By accurately identifying individuals who may be experiencing depression, the model enables the recommendation of necessary support and intervention. The study offers valuable perspectives and a practical tool for clinicians and researchers in the domain of mental health, offering a new avenue for improved mental health assessment and intervention. Harnessing the strength of voice analysis, this approach contributes to enhancing the well-being of individuals affected by depression.
Key Words: Machine Learning; Speech Based Detection; Depression; Convolutional neural network; Mental Illness Prediction
1. INTRODUCTION Mental health conditions affect a substantial portion of the global population, with approximately 450 million individuals experiencing such challenges worldwide. Among these conditions, depressive disorders being a major contributor to the worldwide burden of diseases and are projected to become the second leading cause. Depression, a prevalent and severe mental illness, manifests as persistent feelings of sadness and a lack of enthusiasm in daily activities. Detecting and predicting the presence of mental illness, particularly depression, poses a formidable challenge. Untreated depression can have profound consequences, including diminished motivation, persistent sadness, and low self-esteem. Physical health issues such as weight fluctuations, fatigue, and bodily pains can also arise. Additionally, depression can impair an individual's ability to function effectively, impacting work, education, and the ability to derive pleasure from previously enjoyed activities. Furthermore, individuals with depression may be at an increased risk of developing co-occurring mental health conditions like anxiety and substance abuse. In severe cases, depression can lead to suicidal ideation or attempts. Timely intervention and support are very important when dealing with depression. As the adage goes, "Prevention is better than cure," emphasizing the need and importance of early detection can have far-reaching implications, including escalation of the disorder. The complexity of mental illness makes prediction and identification challenging. Over the past few years, there has been growing interest in utilizing specifically in the realm of supervised machine learning techniques, to discern and forecast mental health conditions based on speech patterns. Supervised learning entails training a model using labelled data to learn parameters that accurately classify audio samples as indicative of a mental illness or not. Subsequently, the model can be deployed to effectively predict mental illness in new audio samples. Such an approach holds promise for delivering more accurate and timely diagnoses, enabling improved treatment and ultimately enhancing the quality of life for individuals grappling with mental health disorders. Supervised learning models based on speech analysis offer an avenue to identify psychological conditions by scrutinizing speech patterns. Individuals with depression exhibit distinctive characteristics, including reduced speaking speed, diminished intonation, lower voice intensity, diminished variations in speech features, and increased pauses. Additionally, alterations in voice bandwidth, amplitude, energy, and other vocal attributes can be observed. Leveraging these features, the model is trained and evaluated, yielding precise and reliable outcomes. In this paper, we delve into the realm of mental illness prediction using supervised learning algorithms and voice analysis. We explore the potential of speech-based models to discern and predict depression, aiming to influence the
© 2023, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 300