A new experimental approach towards the study of structure and dynamic of cognitive processes based on the pacemaker hypothesis about the origin of high-frequency rates of EEG is proposed. To test the hypothesis a new method of micro-structure analysis of brain oscillatory activity has been elaborated. It includes narrow-band frequency filtrating of event-related potentials (ERP), calculation of equivalent current dipoles from the data of multi-channel EEG and their superposition on the structural tomographic magnetic resonance slice of an individual's brain. A new electroencephalographic index of local neural networks activity as frequency selective gamma- or beta-generators characterized by high discreteness of their activity in time and space and imaging pacemaker neurons activity in local networks has been revealed. Frequency selective generators integrate local neural networks into systems due to synchronization of their activity on a common frequency. It happens in case of its phase binding to stimulus and also as a result of periodically appearing bursts of joint activity of many generators working on different frequencies. Maps of spatial localization of equivalent dipole sources of activated frequency selective generators reveal types of interaction of focuses of activity and their changing in the duration of time. When localized dipoles of frequency-selective generators are projected on anatomic magnetic resonance tomograms of individual brain, accuracy of mapping of activated brain areas is increased. Micro-structural analysis of brain oscillatory activity compensateы disadvantages of tomographic methods connected with their low time resolution. Efficiency of the suggested approach in revealing local neural networks activity in human brain is viewed within the framework of research direction earlier established in E.I. Boyko's neurodynamic theory.