Glottal Source Features for Automatic Speech-Based Depression Assessment

Abstract

Depression is one of the most prominent mental disorders, with an increasing rate that makes it the fourth cause of disability worldwide. The field of automated depression assessment has emerged to aid clinicians in the form of a decision support system. Such a system could assist as a pre-screening tool, or even for monitoring high risk populations. Related work most commonly involves multimodal approaches, typically combining audio and visual signals to identify depression presence and/or severity. The current study explores categorical assessment of depression using audio features alone. Specifically, since depression-related vocal characteristics impact the glottal source signal, we examine Phase Distortion Deviation which has previously been applied to the recognition of voice qualities such as hoarseness, breathiness and creakiness, some of which are thought to be features of depressed speech. The proposed method uses as features DCT-coefficients of the Phase Distortion Deviation for each frequency band. An automated machine learning tool, Just Add Data, is used to classify speech samples. The method is evaluated on a benchmark dataset (AVEC2014), in two conditions: read-speech and spontaneous-speech. Our findings indicate that Phase Distortion Deviation is a promising audio-only feature for automated detection and assessment of depressed speech.

Just Add Data: Automated Predictive Modeling and BioSignature Discovery

“…It shields against typical methodological pitfalls in data analysis that lead to overfitting and overestimating performance and therefore to misleading results. The novel machine learning algorithms of JADBIO have been validated by the machine learning and statistical community [10][11][12][13][14][15][16][17][18][19][20] , while the system has produced novel results in nanomaterial property predictions 21 , suicide prediction 22 , speech classification 23 , bank failure prediction 24 , function protein prediction 25 , and breast cancer prognosis and drug response prediction 26 , to name a few.…”

Quantitative comparison of motion history image variants for video-based depression assessment

“…Future work may attempt training and tuning the VGG, testing different versions of the network, assessing the performance of additional classifiers, as well as attempting decision fusion [39] for the different motion representations and feature combinations. Moreover, combining audio-based features [40] could potentially improve overall performance. Testing the two tasks (NorthWind and FreeForm) independently is also of interest.…”

Learning Voice Source Related Information for Depression Detection

“…Depression is known to affect human speech production and cognitive processes: it impacts speech motor control [1,5], and can be identified by prosodic abnormalities and articulatory and phonetic errors [6]. Voice quality has been shown to be affected [7,8,9,10,11] in terms of features such as the shape of the glottal pulse, degree of breathiness, jitter and shimmer. Since depression can sometimes be associated with negative emotions, there have been features motivated from speech emotion recognition research such as [12,13].…”