Exploring Arm Posture and Temporal Variability in Myoelectric Hand Gesture Recognition

Abstract

Hand gesture recognition based on myoelectric (EMG) signals is an innovative approach for the development of intuitive interaction devices, ranging from poliarticulated prosthetic hands to intuitive robot and mobile interfaces. Their study and development in controlled environments provides promising results, but effective real-world adoption is still limited due to reliability problems, such as motion artifacts and arm posture, temporal variability and issues caused by the re-positioning of sensors at each use. In this work, we present an EMG dataset collected with the aim to explore postural and temporal variability in the recognition of arm gestures. Its collection of gestures executed in 4 arm postures over 8 days allows to evaluate the impact of such variability on classification performance. We implemented and tested Stateof-the-Art (SoA) recognition approaches analyzing the impact of different training strategies. Moreover, we compared the computational and memory requirements of the considered algorithms, providing an additional evaluation criteria useful for real-time implementation. Results show a decrease in the recognition of inter-posture and inter-day gestures up to 20%. The provided dataset will allow further exploration of such effects and the development of effective training and recognition strategies.

Robust Real-Time Embedded EMG Recognition Framework Using Temporal Convolutional Networks on a Multicore IoT Processor

“…New machine learning approaches have been extensively explored to enable the design of natural gesture interfaces. They aim to map muscular contraction patterns onto a set of intended gestures [7], using supervised learning methods such as SVM, LDA or ANN [8], [9], [10]. Such approaches reach accuracy above 80% on classifying several hand gestures (from 4 to 12), making them suitable for the design of humanmachine interfaces.…”

“…Nevertheless, the EMG signal is affected by high variability caused by subjects' fatigue, perspiration, changes in the skin-to-electrode interface, user adaptation, and mostly by electrode shifts during multi-day usage [8], [11], [10]. These factors severely limit the long-term usage and the reliability of the EMG-based gesture recognition, leading to an intersessions accuracy drop of up to 30% [12].…”

Online Learning and Classification of EMG-Based Gestures on a Parallel Ultra-Low Power Platform Using Hyperdimensional Computing

“…Furthermore, their computational and memory requirements severely hamper their implementation on resource-constrained platforms. For instance, LDA, which has a training faster than SVM and ANN [16], has time complexity of O(mnt + t 3 ) with a memory footprint of O(mn + mt + nt) memory, where m is the number of samples, n is the number of features and t is defined as min(m, n) [24]. As a result, whereas the number of samples/features grows, the system becomes too resource-hungry to be adapted to a low-power platform.…”

“…it depends on subjectspecific characteristics such as muscular mass, skin thickness, strength of the mean voluntary contractions), and classification algorithms need a training set for each user. Furthermore, EMG setup is intrinsically variable [14]- [16] because of fiber crosstalk, skin perspiration, small movements of the skin-toelectrode interface, power line interference and donning/doffing. As such, small changes of the EMG traces can hinder pattern recognition, degrading the system performance down to unacceptable levels [17].…”

“…An orthogonal approach consists of extending the training dataset, integrating it with samples coming from multiple sessions to gain up to 20% recognition accuracy [16]. A major drawback of increasing the size of the training dataset or repeating the algorithm training lies into its high computational requirements.…”