Rehabilitation through Daily Activities: An Assistive Soft Elbow Exoskeleton for Post-Stroke Patients

Tolstrup, William

4. semester

Robotics, M.Sc.

2024

2024-05-30

65

Slagtilfælde er en hyppig årsag til langvarig funktionsnedsættelse og giver ofte hemiparese (svækkelse i den ene side af kroppen), som gør hverdagsaktiviteter svære. Rehabilitering kan genvinde noget kontrol, men foregår typisk i klinikken med en fysioterapeut eller stationært udstyr; metoder som opgaveorienteret træning og constraint induced movement therapy er ressourcekrævende og ikke velegnede til hjemmet. Dette speciale præsenterer et blødt exoskelet til albueleddet med fokus på komfort ved at være let og diskret. Enheden kombinerer elektromyografi (EMG, muskelaktivitet) og accelerometerdata fra en inertimåleenhed (IMU, bevægelsessensor) for at registrere brugerens intention, og en maskinlæringsmodel (support vector machine, SVM) forudsiger, om brugeren vil bøje eller strække albuen. Både bøjning og strækning assisteres aktivt via et kabledrevet system, der er placeret på brugerens ryg og ført til fastgørelsespunkter på underarmen. I forsøg med to deltagere viste exoskelettet potentiale for at hjælpe i daglige aktiviteter, særligt at løfte en indkøbspose. SVM-modellens nøjagtighed var 82% for den ene deltager og 69% for den anden. Det effektive bevægeudslag var begrænset for begge, og remme kunne glide og reducere effekten. Den gennemsnitlige responstid fra intention til bevægelse var 0,94 sekunder. Samlet set kan prototypen støtte armen og hjælpe med at bøje og strække inden for sine nuværende begrænsninger, men der er tydelige forbedringsbehov. Fremtidigt arbejde bør forbedre respons og pasform, udvide funktionerne til håndled og fingre og teste enheden med post-stroke-patienter i målgruppen.

Stroke is a leading cause of long-term disability and often leaves people with weakness on one side of the body (hemiparesis), making everyday tasks difficult. Rehabilitation can restore some control, but it typically happens in clinics with a therapist or stationary equipment; methods such as task-oriented exercises and constraint induced movement therapy are resource-intensive and not well suited to home use. This thesis presents a soft exoskeleton for the elbow designed for comfort by being lightweight and discreet. The device combines electromyography (EMG, muscle activity) and accelerometer data from an inertial measurement unit (IMU, motion sensor) to detect user intent, and a machine-learning model (support vector machine, SVM) predicts whether the user intends to bend or straighten the elbow. Both flexion and extension are actively assisted via a cable-driven mechanism mounted on the user's back and routed to attachment points on the forearm. In tests with two participants, the exoskeleton showed potential for assisting daily activities, particularly lifting a grocery bag. The SVM reached 82% accuracy for one participant and 69% for the other. The effective elbow range of motion while wearing the device was limited for both, and the straps could slip and reduce effectiveness. The average delay from intended to actual movement was 0.94 seconds. Overall, the prototype can support the arm and assist elbow flexion and extension within its current limitations, but clear improvements are needed. Future work should improve responsiveness and strap fit, expand capabilities to include wrist and finger assistance, and test the device with post-stroke patients, the intended user group.

[This summary has been rewritten with the help of AI based on the project's original abstract]

exoskelet ; SVM ; EMG ; soft exoskelet

Download
View record in AAU Student Projects