Search

Pathological Hand Tremor (PHT) is one of the most prevalent symptoms of some neurological movement disorders such as Parkinson’s Disease (PD) and Essential Tremor (ET). Characterization, estimation, and extraction of PHT is a crucial requirement for assistive and robotic rehabilitation technologies that aim to counteract or resist PHT as an input noise to the system. In general, research in the literature on the topic of PHT removal can be categorized into two major categories, namely, classic and data-driven methods. Classic techniques use hand-crafted features and statistical processing pipelines to model and then extract the tremor while data-driven approaches are trained based on a sizable dataset to allow a computational model (such as neural networks) learn how to estimate the PHT. Since the availability of large datasets, especially in PHT estimation field is a bottleneck, in this feasibility study, we investigate the possibility of combining different recording modalities of PHT to generate a neural network for this purpose. This work explores the potential of jointly using accelerometer data and gyroscope recordings to produce a larger dataset for training a relatively complex network, which can potentially be extended for a deeper generalization.

Recent increases in demand for post-stroke motor rehabilitation services together with limited time of therapist and accessibility issues, in particular for patients living in remote areas, have created a significant burden on healthcare systems worldwide. Semi-autonomous techniques that allow for sharing the time of a therapist between multiple patients have attracted great interest. Among them Learning from Demonstration (LfD) based robots have been studied as solutions to address this growing demand. In this work, a Gaussian Mixture Model (GMM) and Gaussian Mixture Regression (GMR) based LfD approach are proposed to generate a versatile framework to deliver rehabilitation in the absence of the therapist. To collect data for training the models, a bilateral telerehabilitation system is used to enable patient-therapist collaborative task performance is one Degree of Freedom (DOF). The performance and generalizability of the trained model are demonstrated for a variety of patient actions.

The recent evolution of Artificial Intelligence (AI) and deep learning models coupled with advancements of assistive robotic systems have shown great potential in significantly improving myoelectric control of prosthetic devices. In this regard, the paper proposes a novel deep-learning-based architecture for processing surface Electromyography (sEMG) signals to classify and recognize upper-limb hand gestures via incorporation of dilated causal convolutions. The proposed approach has the potential to significantly improve the overall recognition accuracy due to the specific design of the convolutional layers. By using dilated causal convolutions which gradually increases the receptive field of the network, and by applying Conv1D, the proposed architecture eliminates the need for readjustment of the input sequences and inherently takes into account the hidden temporal correlations existing among the available set of sEMG sequences. Contrary to recent hybrid (RNN-CNN) solutions, the proposed architecture neither uses recurrent units nor 2-dimensional convolutions. The publically-accessible NinaPro DB2 dataset is utilized for training and evaluation of the proposed network. Through an extensive set of experiments, it is observed that the proposed architecture significantly outperforms its state-of-the-art counterparts and provides an average 8.71% performance improvement.