I examine in my paper the differences between acoustics of speech parameters of healthy persons and persons diagnosed with depression, in case of read speech. Diagnosis of depression in a proper time may even help saving lives. According to the opinion of psychiatrists, a difference can occur in depressed speech on the basis of hearing, too, as in an acoustic product. Depressed speech is often described by psychiatrists as dull, slow and monotonous. Purpose of my paper is to detect acoustic features of speech, which can effectively separate healthy and depressed groups from each other.
I made recordings for my paper with patients suffering from depression at Department of Psychiatry of Semmelweis Clinic, Péterfy Clinic and Szent János Hospital and with healthy persons keeping the age distributions identical. The participants read a phonetically balanced tale aloud, called “The North Wind and the Sun”. All participants were ranked by a neurologist using a standardized severity scale called BDI . I labeled and segmented the recordings at phoneme level. The selected parameters were obtained by using Praat program.
I used segmental and supra-segmental parameters in the analysis. The segmental parameters such as pitch, first and second formant frequency, jitter, shimmer, HNR, RoT and mel-filter energies were only examined at the phoneme “E”. The supra-segmental parameters were the following: intensity and pitch fluctuation, full pause length, full duration of record, ratio of full pause compared to duration of record, speech rate and articulation rate.
The statistical analysis of acoustic features was carried out using Kolmogorov-Smirnov test and two-patterned T-test and I developed an SVM classification and an SVR regression with cross-evaluation in C#.
Nearly all of the segmental and supra-segmental parameters showed significant differences between the examined groups. As the results are promising, they incite me for further research.