In this paper, we explore the problem of automatic recognition of psychoneurological states: Autism Spectrum Disorders, Down Syndrome, Typical Development of 7–10 years old children from their speech in the Russian language. We described the results of fully automatic recognition based on our proprietary speech dataset. Along with SVM, we used the ComParE features from Computational Paralinguistic Challenges. The results on our dataset showed high performance of automated recognition of psychoneurological states of 7–10 years old children from their speech. The results are theoretically and practically valuable, they will expand the knowledge about human voice uniqueness, possibilities of diagnostics of human psychoneurological states by voice and speech features, and creation of alternative communicative systems.