Experimental results demonstrate multimodal fusion improves biometric authentication and gesture recognition, highlighting the role of EMG and skeleton data.
Key Points
This research aims to enhance biometric authentication and gesture recognition using multimodal data fusion.
Developed a novel dataset combining EMG and 3D hand motion data.
Participants performed three gestures (wave, fist, thumbs-up).
Evaluated classifier performance for authentication and recognition tasks with accuracy metrics.
Unimodal EMG and skeleton classifiers achieved accuracies of 95.6% and 92.5%, respectively.
Classifier-level fusion yielded a 99.4% accuracy for authentication.
Skeleton modality alone reached 99.16% accuracy for gesture recognition.