RADIOELECTRONIC AND COMPUTER SYSTEMS

The article considers the problem of information synthesis for capable of learning control system of prosthetic hand with non-invasive biosignal readout system. Machine learning was carried out within the framework of information-extreme intellectual technology of data analysis, which is based on maximizing the information capacity of the control system in the process of its learning. The article presents a formalized statement of the problem of information synthesis for capable of learning control system of prosthetic hand. A categorical model of machine learning is proposed in the form of a directed graph in which the edge is characterized by the mapping operator of the corresponding set to the other. According to the categorical model, the process of machine learning is considered as a consistent optimization of learning parameters that influence the functional efficiency of the control system. The geometric parameters of the hyperspherical containers of recognition classes and control tolerances on the recognition signs are optimized as learning parameters. The algorithm of machine learning is presented in the form of a two-cycle procedure. The internal cycle of the procedure calculates the geometric parameters of the containers of the recognition classes according to the information criterion, and the outer loop determines the optimal control tolerances for the recognition signs by the maximum value of the criterion. The modified Kulbak information measure was considered as a criterion for optimizing the training parameters. This optimization of control tolerances initially carried out in a parallel procedure when all signs of recognition are changing at every step of training time. Then determined by parallel optimization quasi-optimal control tolerances are used as starting points for their consistent optimization. This approach allows to ensure high reliability and efficiency of machine learning. The hyperspherical deciding rules are constructed as a result of the information-extreme machine learning of control system of prosthetic with non-invasive biosignal readout system and are practically insensitive to the multidimensionality of the space of recognition attributes unlike artificial neural networks. An example of the implementation of the proposed algorithm of machine learning for the recognition of three functional states of prosthesis hand is given.

Farrell, T. R., Weir, R. F. A comparison of the effects of electrode implantation and targeting on pattern classification accuracy for prosthesis control. Biomedical Engineering, IEEE Press, vol. 55, no. 9, 2008, pp. 2198-2211.

Chowdhury, R. H., Reaz, M. B. I., Ali, M. A. B. M., Bakar, A. A. A., Chellappan, K., Chang T. G. Surface Electromyography Signal Processing and Classification Techniques. Sensors, Basel, Switzerland, MDPI Publ., 2013, pp. 12431-12466.

Benatti, S., Farella, E., Benini, L., Gruppioni, E. Analysis of robust implementation of an emg pattern recognition based control. Conference: International Conference on Bioinspired Systems and Signal, Angers, France, BIOSIGNALS, 2014, pp. 45-54.

Rossi, М., Benatti, S., Farella, E., Benini, L. Hybrid EMG classifier based on HMM and SVM for hand gesture recognition in prosthetics. IEEE International Conference on Industrial Technology (ICIT), IEEE Press, 2015, pp. 1700-1705.

Dovbysh, A. S. Osnovy proektuvannya intelektualnykh system [Foundations of the intellectual systems’ designing]. Sumy, Vydavniztvo SumDU Publ., 2009. 172 p.

Dovbysh, A. S., Moskalenko, V. V., Rizhova, A. S. Information-Extreme Method for Classification of Observations with Categorical Attributes. Cybernetics and Systems Analysis, 2016, vol. 52 (2), pp. 224-231