EMG database - EMG Pattern Recognition For Prosthesis Control

<p>This database is one of the results of the MSc dissertation of Adriano de Oliveira Andrade (<a href="http://orcid.org/0000-0002-5689-6606">http://orcid.org/0000-0002-5689-6606</a>)</p> <p>The experimental&nbsp;protocol is fully described in the MSc dissertation (Chapter 5 - pp. 56-60).</p> <p>The avaliable program in R (<a href="https://zenodo.org/api/files/e3190c64-2f7d-4fc6-a070-b0cb2b412156/Import_EMG_Files.R">Import_EMG_Files.R</a>) can be used to import and visualize the EMG data available (<a href="https://zenodo.org/api/files/e3190c64-2f7d-4fc6-a070-b0cb2b412156/EMG-DATA-MSc-AOA.zip?versionId=5216f27e-497a-4e31-8690-a5d5e3b46015">EMG-DATA-MSc-AOA.zip</a>). The EMG data are in in the folders &quot;Isometricos&quot; and &quot;Isotonicos&quot;.</p> <p>&nbsp;</p> <p><strong>MSc dissertation avaiable @</strong></p> <p>[1] Andrade A de O. Metodologia para classifica&ccedil;&atilde;o de sinais EMG no controle de membros artificiais. [manuscrito]. 2000. Uberl&acirc;ndia: Universidade Federal de Uberl&acirc;ndia. doi: <a href="http://doi.org/10.13140/RG.2.2.17314.02242.">https://doi.org/10.13140/RG.2.2.17314.02242.</a></p> <p>[2]&nbsp;http://search.ebscohost.com/login.aspx?direct=true&amp;db=cat08055a&amp;AN=sapere.000058379&amp;lang=pt-br&amp;site=eds-live</p> <p><strong>Abstract -&nbsp;EMG Pattern Recognition For Prosthesis Control</strong></p> <p>One of the major challenges for prosthesis development is to produce devices which mimic their natural counterparts. In general, articial limbs don&#39;t have proper feedback by which the user can assess the status of the prosthesis and the control is very unnatural. Preferably, a subconscious control is desired. Myoelectric control has been widely used as an alternative strategy designed for easier control. However, there is still a lot do be done in order to achieve articial limbs as dextrous as human limbs. In an attempt to contribute to the researches towards better artical limbs, it has been developed an EMG processing system, capable of generate input control to a four degrees of freedom prosthesis. Two major muscle groups (biceps and triceps) were used as source of electromyograc signals, which were discriminated into four different classes: elbow exion, elbow extension, wrist pronation and wrist supination. Those patterns were classied by an articial neural network, which received as inputs the EMG signal features extracted by an autoregressive model. The minimum number of pairs of electrodes and their best positioning for detection, processing and classication were also investigated. To do so, five pairs of electrodes (two on the biceps - long head (B1) and short head (B2) - and three on the triceps - long head (T1), medium head (T2) and lateral head(T3)&nbsp;and one pair of electrodes (on plexo brachial) configuration were considered. Isometric and isotonic contractions were&nbsp;analyzed for each one of those two configurations. The EMG signals were studied in several combinations for each type of contraction. The results show that the configurations using two pairs of electrodes (positioned on B2 and T1) and three pairs of electrodes (positioned on B2, T1 and T2 or B2, T1 and T3), provided accuracy as good as 100%, for the EMG pattern recognition process.</p> <p><strong>Resumo em Portugu&ecirc;s -&nbsp;Metodologia para Classica&ccedil;&atilde;o de Sinais EMG no Controle de Membros Artificiais</strong></p> <p>Um dos grandes desafios atuais das pesquisas envolvendo o aperfei&ccedil;oamento de membros artificiais, &eacute;&nbsp;que esses possam ser controlados de maneira mais natural poss&iacute;vel pelos pacientes. Neste sentido, os processos envolvendo a aquisi&ccedil;&atilde;o e a manipula&ccedil;&atilde;o das informa&ccedil;&otilde;es&nbsp;de controle provenientes do paciente, t&ecirc;m merecido especial aten&ccedil;&atilde;o. Dentre as diversas t&eacute;cnicas de controle poss&iacute;veis, uma das que tem alcan&ccedil;ado melhores resultados utiliza a atividade eletromiogr&aacute;fica resultante de contra&ccedil;&otilde;es volunt&aacute;rias de determinados grupos musculares. Numa tentativa de contribuir para aquelas pesquisas, foi desenvolvido um sistema de processamento de sinais eletromiogr&aacute;ficos (EMG), capaz de fornecer entradas de controle para uma pr&oacute;tese com quatro graus de liberdade. Para tal, sinais EMG provenientes dos grupos musculares tr&iacute;ceps e b&iacute;ceps foram classificados em quatro padr&otilde;es distintos: flex&atilde;o e extens&atilde;o do cotovelo, prona&ccedil;&atilde;o&nbsp;e supina&ccedil;&atilde;o do punho. A classifica&ccedil;&atilde;o dos padr&otilde;es foi feita atrav&eacute;s de uma rede neural artifical que recebe como entrada as caracter&iacute;sticas dos sinais eletromiogr&aacute;ficos, extra&iacute;das atrav&eacute;s de um modelo autoregressivo. Outro objetivo desta pesquisa foi buscar o n&uacute;mero m&iacute;nimo de pares de eletrodos e os s&iacute;tios mais adequados para a detec&ccedil;&atilde;o, processamento e classifica&ccedil;&atilde;o satisfat&oacute;ria dos movimentos executados. Foram feitas an&aacute;lises considerando 5 pares de eletrodos, sendo dois sobre o b&iacute;ceps - na cabe&ccedil;a longa (B1) e na cabe&ccedil;a curta (B2) - e tr&ecirc;s sobre o tr&iacute;ceps - na cabe&ccedil;a longa (T1), na cabe&ccedil;a medial (T2) e na cabe&ccedil;a lateral (T3); e um par de eletrodos sobre o plexo braquial. Os experimentos foram realizados considerando-se contra&ccedil;&otilde;es isom&eacute;tricas e isot&ocirc;nicas. Aqueles sinais foram analisados em diversas combina&ccedil;&otilde;es, para cada tipo de contra&ccedil;&atilde;o, numa tentativa de se encontrar aquela que apresentasse melhores resultados. Os resultados mostraram que as combina&ccedil;&otilde;es envolvendo o uso de dois pares de eletrodos posicionados sobre os s&iacute;tios B2 e T1; e tr&ecirc;s pares de eletrodos posicionados sobre os s&iacute;tios B2, T1 e T2 ou B2, T1 e T3 apresentaram melhores performances, com taxas de acerto de at&eacute; 100%.</p>