Аннотация:Despite of there are classic methods for human functional state specification (such as EEG, ECG, EMG, CGR et cetera), each of them has some disadvantages. Therefore, new physiological correlates of human state changes are still required. Speech signal seems to be representative since speech changes together with condition and practical as far as voice recording is non-invasive and requires no contact sensor. Acoustic speech characteristics (like formants, fundamental frequency and speech rate) indicate emotional state and the degree of fatigue. Another type of speech analysis is non-linear which consider, for instance, correlation dimension (D 2). This method is founded on Taken's theorem and allows reconstruct phase space trajectory based on the values of one variable, taken with time lag. D 2 could be calculated for signal of different length. Presumably, D 2 for phrase reflects mental processes stability and D 2 for words or other short segment reflects autonomic system state. Non-linear characteristics of healthy human and human with voice tract pathology are different. Moreover, using them it is possible to determine emotional state. This research area seems to be very promising but insufficiently developed yet. It is important to mark different kinds of experiment design. There are various way to get speech data such as asked participants to read sentences or to react on stimulus and this will influence on results. Nowadays the patterns of non-linear parameters changes need confirmation by classic physiological methods. Researches in this way enable to clarify the relationship between speech function and the human functional state and, on the next stage, correlate speech parameters dynamics with neurophysiological changes.