Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Watch this Oxford seminar where Samuel George-White presents research on using artificial intelligence to enhance human-machine interaction through neural intent decoding from non-invasive muscle signals. The presentation explores how high-density electromyography (HD-EMG) and EEG data were collected during a GO/NO-GO task to capture movement preparation and cancellation signals. Learn about the data preprocessing techniques employed, including filtering, principal component analysis for dimensionality reduction, and normalization. The speaker discusses the convolutional neural network developed for the study, which achieved 72.4% accuracy in two-class classification and 49.6% in three-class classification of movement-related brain states. Discover how visualization methods like GradCAM and temporal sliding windows were used to interpret model decisions and track neural signal evolution over time. The findings demonstrate the potential of surface EMG for decoding brain activity, with promising applications for more intuitive control systems in prosthetics, rehabilitation, and real-time assistive technologies.
