The effective models of human-computer interaction by natural form are investigated. The developed Russian speech recognition system SIRIUS is combined with other natural modalities processors (movements of head, lips, etc.) for creation of multimodal interfaces. In contrast to English the Russian language has much more variety on word-form level and so the size of recognized vocabulary sharply increases as well as quality and speed of the processing decrease. To avoid these problems the additional level of speech representation (morphemic level) was applied. The system ICANDO (Intellectual Computer Assistant for Disabled Operators) intended for assistance to persons without hands or with disabilities of their hands or arms for human-computer interaction combines the modules for automatic speech recognition and head tracking in one multimodal system.The applications for telecommunication, biomonitoring and voice operating for disabled peoples are realized during EU projects.