N.N. Lebedeva, E.D. Karimova, E.A. Kazimirova

Speech analysis can give a lot of information about the human condition. Sometimes, the voice signal is the only available source of information about the human condition. Analysis of voice signal data is used in medical psychology, for identification and verification of the speaker, in psychiatry and neurology. Diagnosis by speech signal has important methodological advantages - it is non-invasive, non-contact, and, if necessary, unknown to the subject. Thirteen volunteers (6 male, 7 female) were recruited in the study of voice analysis. All participants read out the same questions. Vowels were separated from the records and processed by the original program "Radiometer". So, the non-linear characteristics of the speech signal - fractal dimension D2 - were obtained. These parameters for each vowel were compared with the results of Psychophysiology tests and EEG data. To estimate the volunteer functional state we conducted EEG-recording with eyes closed and eyes opened, ECG-recording, psychological tests (health, activity, mood and anxiety definition). All results were compared by such nonparametric statistical methods as Mann-Whitney criterion; the correlative correspondences were also detected. We found out some correlation between D2 and individual characteristics of the person. The participants were divided into groups based on gender or on the power of the alpha rhythm with closed eyes. The dependence of the D2 with the EEG parameters and the D2 with the psychophysiology parameters can be different for each group. For example, the D2 values of women increase with reactive anxiety growth, and for men these two quantities are independent. In addition, there was found that the value of D2 is affected by word-s semantics. Specifically, the D2 value increase in reading a key part of the issue.