RESUMEN
This systematic review aims to assess and summarize the current landscape in exoskeletons and orthotic solutions developed for upper limb medical assistance, which are partly or fully produced using 3-dimensional printing technologies and contain at least the elbow or the shoulder joints. The initial search was conducted on Web of Science, PubMed, and IEEEXplore, resulting in 92 papers, which were reduced to 72 after removal of duplicates. From the application of the inclusion and exclusion criteria and selection questionnaire, 33 papers were included in the review, being divided according to the analyzed joints. The analysis of the selected papers allowed for the identification of different solutions that vary in terms of their target application, actuation type, 3-dimensional printing techniques, and material selection, among others. The results show that there has been far more research on the elbow joint than on the shoulder joint, which can be explained by the relative complexity of the latter. Moreover, the findings of this study also indicate that there is still a gap between the research conducted on these devices and their practical use in real-world conditions. Based on current trends, it is anticipated that the future of 3-dimensional printed exoskeletons will revolve around the use of flexible and high-performance materials, coupled with actuated devices. These advances have the potential to replace the conventional fabrication methods of exoskeletons with technologies based on additive manufacturing.
RESUMEN
Deep learning (DL) models are very useful for human activity recognition (HAR); these methods present better accuracy for HAR when compared to traditional, among other advantages. DL learns from unlabeled data and extracts features from raw data, as for the case of time-series acceleration. Sliding windows is a feature extraction technique. When used for preprocessing time-series data, it provides an improvement in accuracy, latency, and cost of processing. The time and cost of preprocessing can be beneficial especially if the window size is small, but how small can this window be to keep good accuracy? The objective of this research was to analyze the performance of four DL models: a simple deep neural network (DNN); a convolutional neural network (CNN); a long short-term memory network (LSTM); and a hybrid model (CNN-LSTM), when variating the sliding window size using fixed overlapped windows to identify an optimal window size for HAR. We compare the effects in two acceleration sources': wearable inertial measurement unit sensors (IMU) and motion caption systems (MOCAP). Moreover, short sliding windows of sizes 5, 10, 15, 20, and 25 frames to long ones of sizes 50, 75, 100, and 200 frames were compared. The models were fed using raw acceleration data acquired in experimental conditions for three activities: walking, sit-to-stand, and squatting. Results show that the most optimal window is from 20-25 frames (0.20-0.25s) for both sources, providing an accuracy of 99,07% and F1-score of 87,08% in the (CNN-LSTM) using the wearable sensors data, and accuracy of 98,8% and F1-score of 82,80% using MOCAP data; similar accurate results were obtained with the LSTM model. There is almost no difference in accuracy in larger frames (100, 200). However, smaller windows present a decrease in the F1-score. In regard to inference time, data with a sliding window of 20 frames can be preprocessed around 4x (LSTM) and 2x (CNN-LSTM) times faster than data using 100 frames.
