Stress is a term used to describe emotional distress. Our own mental health as well as that of those around us may be impacted. While anxiety is a common, potentially terrifying reaction to stress, it can also result in panic attacks. The most recent figures show that millions of people around the world experience one or more mental disorders1. Early diagnosis of mental illness can help with both disease progression and treatment in general.Effective feature detection is made possible by convolutional neural networks. A convolutional neural network is particularly good at identifying images because of the way that simple features that are discovered early in the network (like lines) may be combined to create more complex features (like the straightforward curve lines that make up a left eye). Audio datasets are used to train and assess classification algorithms like CNN. After using acoustic feature extraction as pre-processing, CNN is employed to accurately identify the audio based on emotions. This enables us to predict whether the person is anxious or not.This study reviews machine learning-based mental disease detection system developments. Early detection and management can improve patient outcomes for mental illness, a global issue. Speech, text, and physiological data train detection systems, according to the paper. Deep learning and support vector machines are also reviewed. The paper discusses data bias and interpretability issues in mental disease detection systems. Finally, these systems may improve mental health care results and save cost.