SYSTEM FOR READING AND ANALYZING SIGNALS OF THE HUMAN BRAIN
The given work is devoted to the methods for analysis of signals of the encephalograph by neural networks. A system for analysis of read brain signals using a handheld computer Raspberry Pi 3b+ is proposed. On the basis of this system it is proposed to create a neural network which will analyze the received signals. Analytical review of a device for reading and analysis of human brain signals using an encephalograph was performed. The parameters of the components were calculated and described the necessary components. The existing methods of receiving signals of the human brain are analyzed. Methods of analysis for additional algorithms of machine analysis are looked at.
The basis of Neurocomputer Interface is the recognition of patterns of brain biopotentials. If the subject can change the nature of his biopotentials, for example, by performing certain mental tasks, the NCI system could translate these changes into control codes, such as by moving the mouse cursor on a computer screen or the hand of a robot manipulator. You can also use these codes to select letters on a "virtual keyboard" or to control a wheelchair.
The most of current Neurocomputer Interfaces are preprogrammed. Firstly, that means that it is not consider brain characteristics of each user. This causes difference in work for each person so it may be very easy for one user and really hard for other. Secondly, a device can do built in tasks only. It can be really hard and even impossible to add new functionality to device.
These restrictions can be removed by a device that can collect signals of human brain, label them and create the model that represents their connections and then use this model to recognize current state of human brain. This problem can be solved by connecting neurointerface to the pocket PC Raspberry pi 3b+ because it is compact enough and already have all needed data interfaces on RPi bus. Although, it can use python programming language that allows us to use tensorflow package that have already have a lot of functionality in using of neural networks.