Paper Title
Depression Level Detection

Abstract
The affective computing community and the area of artificial intelligence are becoming increasingly interested in building digital systems to assist physicians in better determining the degree of a person's depression. The voice and video applications are beneficial for treating depression. Nonetheless, understanding of manual design and domain is still crucial in picking the role that makes the process labour intensive and subjective. Deep-learned applications based on neural networks have outperformed hand-crafted apps in a variety of fields in recent years.In this post, we offer deep-learned programmes that can effectively detect the extent of voice and facial sadness and fix the aforementioned concerns. Convolutional Neural Networks (CNN) are initially created in the suggested approach for learning deep-learned features from spectrograms and descriptive raw waveforms. Alternatively, the cutting-edge texture descriptors known as median robust extended local binary patterns are manually extracted from spectrograms. They propose using fine-tuning layers to integrate the CNN raw and spectrogram to improve depression detection efficiency by capturing complementing information inside handcrafted features and deep-learned functionality.In addition, to address the issue of small sample size, a data-augmentation strategy was adopted. Experiments on depression libraries indicate that, when compared to cutting-edge audio-based approaches for diagnosing depression, our strategy is more reliable and efficient. Keywords - Depression Voice communication, photos, automatic recognition Median generalized Robust Local Binary Patterns (MRELBP).