Depression is a common and serious medical illness that negatively affects how you feel, the way you think and how you act. Automatic detection of depression is state-of-the-art problem which and its oriented towards objectifying the diagnosis procedure. Daniela Janeva demonstrates two different approaches in her Diploma Thesis – detection of depression based on speech prosody and EEG signals (https://feit.ukim.edu.mk/en/computer-…).
Speech prosody is concerned with those elements of speech that are not individual phonetic segments (vowels and consonants) but are properties of syllables and larger units of speech, including linguistic functions such as intonation, tone, stress, and rhythm. For this system, audio files of depressed and control individuals were used. Spectrograms were generated for 2s audio segments, and classification was done with convolutional neural networks and spectrograms as input. The results are promising for further development of the system.
Brain activity is observed using an Electroencephalogram. The registered abnormalities in the brain waves are biomarkers for diagnosis of many diseases and disabilities. The characteristics of the EEG signals are a solid base for developing a system for detecting depression. For that goal, different machine learning algorithms were trained and evaluated. The best results were achieved using the Random Forest algorithm. The developed GUI provides easier access to the EEG files and shows classification results on the screen.
Student: Daniela Janeva
Mentor: Branislav Gerazov
