Биомедицинская инженерия и электроника
Электронный научный журнал

Технические науки
METHOD OF FOREARM PROSTHETICS USING ARTIFICIAL NEURAL NETWORKS
Kuzovik V.D. 1, Yakovenko D.K. 2, Monchenko О.V. 1, Tishchenko E.A. 1

1. National Aviation University
2. WePlay Esports

Abstract:

Loss of limb functionality creates significant discomfort for a person, significantly affects the quality of his life. Today, the improvement of myoelectric prostheses, which are the most modern, can eliminate the existing shortcomings and improve the life of their users in general. The paper proposes a method of modification of the block of processing of electromyographic signals by means of a neural network. An important aspect of the decision to use neural networks for bioelectrically controlled prostheses is their versatility and simplification of teaching the patient to use the prosthesis in household chores. As of 2021, neural networks have developed very strongly and covered each of the areas. Their teaching method is not too difficult and does not require unnecessary intervention. However, there is an opposite side, which consists in the collection of material for the training of this network. Age, weight, height and size of the stump - all these aspects affect the value of the original electromyographic signals, so it requires a large base and description of each of the signals. Nowadays myoelectric prostheses are the most advanced, they convert muscle activity into information, which is processed by the processor and gives a signal to electric motors that are used to control the movements of artificial limbs. Because the battery and electronic motors are used, no bulky belts or harnesses are required to operate the myoelectric prosthesis. The main disadvantages of this type of prosthesis now are their weight and cost. Their large weight is primarily due to the fact that the myoelectric prosthesis contains a battery and electric motors, and, unlike prostheses that are driven by the body, does not use any harnesses to balance the weight throughout the body. Another disadvantage of myoelectric prostheses is their cost. Although it is currently more expensive than other types of prostheses, it also offers better quality in terms of both cosmetics and functionality. The proposed control system of the bioelectric prosthesis with the module of the positional perception device allows specialists in the natural mode to obtain information about the position of the fingers of the hand, the beginning of the movement, the end of the movement and the process of movement. The positive effect of the method: it provides ease of use of the prosthesis in self-care, increases the accuracy of coordination of the positional position of the fingers and dosing force, thereby reducing the time of work operations when performing targeted actions. Since the system works only for receiving and processing the electromyographic signal, the main modified unit will be an electronic control unit for the movement of the fingers of the artificial brush. It is responsible for deciding, what input data will be taken by the following elements in the circuit. The process of choosing the best answer is complex, which means that the main part of the price of the final product is fixed by the decision-making algorithm. That is why the use of a neural network in this unit is reasonable in terms of price and availability, because the algorithm of operation and training of the network will not differ from the state, age and sex of the person. Further training of the neural network can be divided into several stages, namely finding the results of neurons, calculating the result of error by epoch, finding errors in each layer, changing the weights between neurons and then storing weights in a connected database. Thus, a method was developed to help solve the current shortcomings of active myoelectric prostheses, namely the price and affordability for the general public. The materials of the work can be used in the practical activities of specialists working in the medical and biomedical fields.

Keywords: prosthesis, electromyographic signal, neural network, myographic prosthesis, software, sensor circuit

Сайт работает на RAE Editorial System