Development of a robust myoelectric control architecture for lower limb robotic prostheses applications.

Garikayi, Talon (2018-12)

Thesis (PhD)--Stellenbosch University, 2018.

Thesis

ENGLISH ABSTRACT: Traumatic events such as accidents or vascular and circulatory disorders often lead to amputation of the lower limb. To increase mobility most amputees are tted with a passive prosthetics. However, the use of a passive foot with a xed ankle has short term e ects, such as asymmetric gait, increased muscle contraction on the intact side and higher metabolic energy expenditure. The long-term e ects are osteoarthritis, osteoporosis, back pain and to a large extent musculoskeletal problems. As a result, arti cial prosthetic limbs are regarded by the amputees as exotic lifeless attachments to the body and not as a non-biological extension of the human body. Mechatronic systems coupled with intelligent control architectures provide the platform to restoring an amputee's overall mobility related lifestyle. However, the recovered gait is largely in uenced by the extent of amputation and functional level of the prosthesis. The transtibial osteomyoplastic amputation technique offers residual muscles that are active throughout the gait cycle. These muscles offer potential sites for extracting surface electromyography (sEMG) signals. The study presents a novel methodology which seeks to utilise these residual signals to control an artificial limb by predicting the human movement intentions. A protocol was developed for the acquisition and analysis of electromyography signals from the identified muscles. The available SENIAM and ISEK standards were found to be insuficient during the recording of signals from the residual stump as some of the anatomical landmarks were missing. The Soleus muscle responsible for plantarflexion was not accessible on the residual limb thereby providing challenges on using the SENIAM standards for selecting a muscle for the plantarflexion movement. Tibialis anterior, Medialis Gastrocnemius and Lateralis Gastrocnemius muscles were able to provide sEMG signals with sufficient signal properties for developing a myoelectric pattern recognition architecture. The main goal was to develop a robust intelligent control system architecture for a robotic prosthetic lower limb capable of enhancing human mobility with great stability. The functionality of a robotic limb is highly governed by kinetics, kinematics and the dynamics of the mechanical structure when interfaced with the human body. Therefore, the structure and parameters of the actuation model for complex joint angle prediction and an intuitive neural interface mechanism for intention detection were developed based on experimental results from biomechanics experiments. A pattern recognition algorithm was developed based on 23 signal features. Principal component analysis was used for dimensionality reduction on the extracted feature set. A total of 22 classifiers were tested and the Linear Support Vector Machine produced an average of 100% classification accuracy on training data with 20% of the training data being reserved for validation. The intelligent architecture produced an average of 99.25% classification accuracy on new unlabelled test data. The system was optimised using force sensitive resistors to detect heel strike, toe o and beginning of the swing and stance phases of gait. A dual inertial measurement system was used to predict the position of the limb in space thereby providing feedback on limb performance to the main controller. The use of adaptive lters on signal acquisition improved signal quality and the use of Kalman lters on feedback sensors provided a robust system which was able to achieve the desired control objective even in the event of partial or missing input signal as they predicted the intended signal based on the previously correct signal input. This study revealed that the concept developed has the potential to improve the lives of many amputees as it has the ability to restore normal gait to the satisfactory level of the amputee. The intuitive control of the prosthetic limb provided by the sEMG signals and the inertial sensor feedback system minimises the need for the situational attentiveness of the amputee with regards to the operation of the powered prosthetic.

AFRIKAANSE OPSOMMING: Traumatiese gebeurtenisse soos ongelukke of vaskulêre en bloedsomloopafwykings lei dikwels tot amputasie van die onderste ledemaat. Om die mobiliteit te verhoog, word die meeste geamputeerde toegerus met 'n passiewe prostese. Die gebruik van 'n passiewe voet met 'n vaste enkel het kort termyn effekte soos asimmetriese gang, verhoogde spierkontraksie op die ongeskonde kant en hoër metaboliese energieverbruik. Die langtermyn-effekte is osteoartritis, osteoporose, rugpyn en tot 'n groot mate muskuloskeletale probleme. Gevolglik word kunsmatige ledemate deur die geamputeerde beskou as eksotiese lewenslose aanhangsels binne die liggaam en nie as 'n nie-biologiese verlenging van die menslike liggaam nie. Megatroniese stelsels, tesame met intelligente beheer-argitekture, bied die platform om 'n geamputeerde se algehele mobiliteitsverwante leefstyl te herstel. Die herstelde gang is egter grootliks beïnvloed deur die mate van amputasie en funksionele vlak van die prostese. Die transtibiale osteomyoplastiese amputasietegniek bied oorblywende spiere wat aktief is in die loopsiklus. Hierdie spiere bied potensiële areas vir die opneem van oppervlak-myo-elektroniese seine. Die studie bied 'n nuwe metodologie aan wat daarop gemik is om hierdie residuele seine te gebruik om 'n kunsmatige ledemaat te beheer deur die menslike bewegings bedoelings te voorspel. 'n Protokol is ontwikkel vir die verkryging en analise van elektromyografiese seine van die geïdentifiseerde spiere. Die beskikbare SENIAM- en ISEK-standaarde was onvoldoende tydens die opname van seine van die oorblywende stomp, aangesien sommige van die anatomiese landmerke ontbreek. Die Soleus-spier wat verantwoordelik is vir plantarfieksie, was nie toeganklik op die oorblywende ledemaat nie en bied dus uitdagings om die SENIAM-standaarde te gebruik om 'n spier vir die plantarfleksie-beweging te kies. Tibialis anterior, Medialis Gastrocnemius en Lateralis Gastrocnemius spiere was in staat om sEMG seine te voorsien met voldoende sein eienskappe vir die ontwikkeling van 'n myoektriese patroon herkenning argitektuur. Die hoofdoel was om 'n robuuste intelligente beheerstelselargitektuur vir 'n robotprotese-onderste ledemaat te ontwikkel wat die menslike mobiliteit met hoë stabiliteit kan verbeter. Die funksionaliteit van 'n robot-ledemaat word sterk beheer deur kinetika, kinematika en die dinamika van die meganiese struktuur wanneer dit met die menslike liggaam inmeng. Die struktuur en parameters van die aktuasie model vir komplekse gesamentlike voorspelling en 'n intuïtiewe neurale koppelvlak meganisme vir voornemende opsporing is daarom ontwikkel op grond van eksperimentele resultate uit biomeganiese eksperimente. 'n Patroonherkenningsalgoritme is ontwikkel op grond van 23 seinmerke. Die hoof komponent analise is gebruik vir dimensionaliteitsreduksie op die opgeneemde stel seine. 'n Totaal van 22 klassifikasie algoritmes is getoets en die Lineêre Ondersteuningsvektormasjien het gemiddeld 100% klassifikasie akkuraatheid op opleidingsdata geproduseer met 20% van die data wat vir validering gereserveer is. Die intelligente argitektuur het 'n gemiddeld van 99,25% klassifikasie akkuraatheid op nuwe ongemerkte toetsdata opgelewer. Die stelsel is geoptimaliseer deur gebruik te maak van kraggevoelige weerstande om hakstaking, tone-af en begin van die swaai- en houdingsfases van die gang te bepaal. 'n Dubbele traagheidsmetingsisteem is gebruik om die posisie van die ledemaat in die ruimte voor te stel en sodoende terugvoer te gee aan ledemate-prestasie aan die hoofbeheerder. Die gebruik van adaptiewe filters op sein verkryging verbeter sein kwaliteit en die gebruik van Kalman-filters op terugvoer-sensors het 'n robuuste stelsel gelewer wat in staat was om die verlangde beheerdoelwit te behaal, selfs in die geval van gedeeltelike of ontbrekende insetsein, aangesien hulle die beoogde sein voorspel het gebaseer op die voorheen korrekte seininvoer. Hierdie studie bewys dat die konsep wat ontwikkel is die potensiaal het om die lewens van baie geamputeerdes te verbeter aangesien dit die vermoë het om normale gang te herstel tot die bevredigende vlak van die geamputeerde. Die intuïtief beheer van die prostetiese ledemaat wat deur die elektromyografiese seine en die traagheidssensor terugvoerstelsel verskaf word, verminder die behoefte aan die situasionele aandag van die geamputeerde ten opsigte van die werking van die aangedrewe prostese.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/105115
This item appears in the following collections: