Analytical review of automatic systems for depression detection by speech

In recent years the interest in automatic depression detection has grown within medical and scientific-technical communities. Depression is one of most widespread mental illnesses that affects human life. this review we present analyze latest researches devoted to detection. Basic notions related definition were specified, includes both unimodal multimodal corpora containing records informants diagnosed with control groups non-depressed people.
 Theoretical practical which automated systems for reviewed. The last ones include as well systems. A part reviewed addresses challenge regressive classification predicting degree severity (non-depressed, mild, moderate severe), another solves a problem binary presence (if person depressed or not). An original methods computing informative features three communicative modalities (audio, video, text information) presented. New every modality all total are defined. popular studies neural networks. survey shown main psychomotor retardation strong correlation affective values valency, activation domination, also there been observed an inverse between aggression. Discovered correlations confirm interrelation disorders emotional states. trend many papers combining improves results