Robotic hand control using support vector machine and open source electromyograph

Trg elektrotehničnih izdelkov je velik in raznolik. Ponuja tudi majhne računalnike v velikosti bančne kartice (zgled: Raspberry Pi) ali mikrokrmilnike (zgled: Arduino), ki so cenovno dostopni, odprto kodni in preprosti za uporabo. Hkrati je na voljo malo cenovno dostopnih in odprtokodnih naprav, ki bi bile uporabne za klasificiranje in analizo človeških signalov, kakršen je elektromiograf (EMG). Želeli smo preizkusiti napravo, ki bi lahko zapolnila zgoraj omenjeno vrzel. Preizkusili smo petkanalni EMG ščit za Arduino na treh zdravih moških, starosti 15, 22 in 27 let (starost = 21 ± 6(SD) let). Njihova naloga je bila, da izvajajo preproste gibe s prsti, na podlagi katerih smo naučili EMG klasifikator gibov prstov. Klasifikacija gibov je bila osnovana na metodi podpornih vektorjev. Klasifikator je skupaj z EMG Arduinom dosegel povprečno točnost 78 % pri prepoznavanju ustreznih gibov prstov in posledično aktiviral ustrezni prst robotske roke. Naši rezultati potrjujejo, da lahko z naučenim klasifikatorjem gibov, ki temelji na metodi podpornih vektorjev, vodimo robotsko roko v realnem času z uporabo Arduina. To nam je uspelo z uporabo cenovno dostopne in prilagodljive naprave, ki ima potencial kot izobraževalni pripomoček na področju analize in uporabe človeških signalov.Off-the-shelf electronic market is large and diverse. It includes credit card size computers (example: Raspberry Pi) or microcontroller boards (example: Arduino) that are relatively low cost, open source and easy to use. Nevertheless, there is a lack of off-the-shelf, open source devices that would enable us to learn about and make use of physiological signal processing. An example of such a device is an electromyograph (EMG). In this thesis, we investigated if an EMG device could fulfil the afore mentioned gap. EMG device was a five channel open source Arduino EMG shield. The performance of the device was evaluated on three healthy male subjects aged 15, 22, 27 (age = 21 ± 6(SD) years). They were instructed to perform simple finger movements, which we classified and replicated on the robotic hand. The EMG signal classification was performed using a support vector machine (SVM) algorithm. In our experimental setup, the average EMG signal classification accuracy was 78 %. Our results confirm, that by using a trained movement classifier based on the support vector machine algorithm we can control a robotic hand in real time by utilising the EMG signal acquired by an Arduino EMG shield. We achieved our results by using a cost effective and customisable device holding the potential to provide access to easier human-machine interface prototyping and learning about neurophysiology