Statistical analysis of the acoustical parameters of depressed speech

OData support
Supervisor:
Dr. Vicsi Klára
Department of Telecommunications and Media Informatics

Human speech is a very complex phenomenon, affected by many factors. These factors are, for example, sex, age health and emotional state of the speaker. These kind of features are percepted by people through a complex hearing mechanism. However, it is important to examine human speech in its depth, as an acoustic product, because, if we are able to recognize a disorder of a vocal organ, or a mental disease in time, it can even save lives.

I examine in my work, what kind of differences can be experienced in speech of people suffering from depression, compared to normal speech. As a first step, it was necessary to collect voice patterns of people suffering from depression, and to their voice-levelled segmentation. I made the acoustic measures with the help of a program named Praat, and I compared the gained results to voice records of the Hungarian Reference Database (MRBA). My examinations were focused to jitter, shimmer, and harmonics-to-noise ratio and to mel-frequenced spectral steepness, using two wovels („e” and „a”) both for male and female speakers.

I detected with the help of statistical methods, that there are significant differences in voice patterns of depressed people. These results can be useful in a later developed statistical classification process, which would be used to determine from the speech, whether it is necessary to give a medical help to the patient to his/her recovery, or not.

Downloads

Please sign in to download the files of this thesis.