Rim Barioul

The prediction of gestures in shared public spaces is important not only to ensure good functionality of agents but also the safety of humans and their acceptance of the hybrid society concept. In this context, surface electromyography (sEMG) combined with inertial measurement unit (IMU) signals is investigated in this work to predict dynamic hand gestures. Our proposed algorithm is designed to make predictions, while these signals are being observed, combining a Nadaraya–Watson kernel estimator and an entropy-based decision function. We not only achieve a high accuracy, but also a significant earliness of prediction.

Gestures are one of the most intuitive ways that humans use to interact with others or convey information. The idea that hand gestures could facilitate human–machine interaction has recently gained increasing interest among researchers. Various technologies have been investigated, providing both visual and sensor-based gesture recognition. While camera-based solutions suffer from the constraint of specific and expensive laboratories, wearable sensor-based solutions allow lower costs and higher flexibility, enabling gesture recognition even in public spaces. Although several solutions are available in the literature, most of them focus on specific sensor principles and specific gestures. The aim of this work is to recognize basic gesture components, defined as primary elements that compose more complex gestures, using both force myography (FMG) and electromyography (EMG), and to highlight their strengths and weaknesses. This will provide the foundation for the recognition of more complex human upper limb movements. To this end, a laboratory study was conducted with ten participants. FMG signals were collected by means of a wearable sensor network consisting of an instrumented smart band with eight pressure sensors and a wireless datalogger. EMG data were acquired using three commercial sensors. The recorded data were analyzed using k-nearest neighbor classifier and extreme learning machine algorithms. The results showed that the data recorded using FMG had higher accuracy in recognizing the ten different static hand gestures studied compared to the EMG data.

Gesture recognition plays a critical role in applications ranging from gaming and remote manipulation to assistive technologies for individuals with disabilities. Force myography (FMG) has recently emerged as a promising approach for gesture recognition due to its strong classification capabilities. In this study, eleven American Sign Language (ASL) gestures were classified using FMG signals. To improve classification accuracy, three advanced swarm-based feature selection algorithms: the Binary Gray Wolf Optimizer (BGWO), Binary Grasshopper Optimization Algorithm (BGOA), and Binary Hybrid Gray Wolf Particle Swarm Optimizer (BGWOPSO), were evaluated as wrapper methods for an Extreme Learning Machine (ELM) classifier. Ten volunteers contributed data using a bracelet equipped with six force-sensitive resistor (FSR) sensors. Results show that the BGWOPSO-based feature selection approach significantly outperformed the other methods, increasing ELM classification accuracy from 55.82% to 91.36%.

Effective feature extraction and selection are crucial for the accurate classification and prediction of hand gestures based on electromyographic signals. In this paper, we systematically compare six filter and wrapper feature evaluation methods and investigate their respective impacts on the accuracy of gesture recognition. The investigation is based on several benchmark datasets and one real hand gesture dataset, including 15 hand force exercises collected from 14 healthy subjects using eight commercial sEMG sensors. A total of 37 time- and frequency-domain features were extracted from each sEMG channel. The benchmark dataset revealed that the minimum Redundancy Maximum Relevance (mRMR) feature evaluation method had the poorest performance, resulting in a decrease in classification accuracy. However, the RFE method demonstrated the potential to enhance classification accuracy across most of the datasets. It selected a feature subset comprising 65 features, which led to an accuracy of 97.14%. The Mutual Information (MI) method selected 200 features to reach an accuracy of 97.38%. The Feature Importance (FI) method reached a higher accuracy of 97.62% but selected 140 features. Further investigations have shown that selecting 65 and 75 features with the RFE methods led to an identical accuracy of 97.14%. A thorough examination of the selected features revealed the potential for three additional features from three specific sensors to enhance the classification accuracy to 97.38%. These results highlight the significance of employing an appropriate feature selection method to significantly reduce the number of necessary features while maintaining classification accuracy. They also underscore the necessity for further analysis and refinement to achieve optimal solutions.

The recognition of hand signs is essential for several applications. Due to the variation of possible signals and the complexity of sensor-based systems for hand gesture recognition, a new artificial neural network algorithm providing high accuracy with a reduced architecture and automatic feature selection is needed. In this paper, a novel classification method based on an extreme learning machine (ELM), supported by an improved grasshopper optimization algorithm (GOA) as a core for a weight-pruning process, is proposed. The k-tournament grasshopper optimization algorithm was implemented to select and prune the ELM weights resulting in the proposed k-tournament grasshopper extreme learner (KTGEL) classifier. Myographic methods, such as force myography (FMG), deliver interesting signals that can build the basis for hand sign recognition. FMG was investigated to limit the number of sensors at suitable positions and provide adequate signal processing algorithms for perspective implementation in wearable embedded systems. Based on the proposed KTGEL, the number of sensors and the effect of the number of subjects was investigated in the first stage. It was shown that by increasing the number of subjects participating in the data collection, eight was the minimal number of sensors needed to result in acceptable sign recognition performance. Moreover, implemented with 3000 hidden nodes, after the feature selection wrapper, the ELM had both a microaverage precision and a microaverage sensitivity of 97% for the recognition of a set of gestures, including a middle ambiguity level. The KTGEL reduced the hidden nodes to only 1000, reaching the same total sensitivity with a reduced total precision of only 1% without needing an additional feature selection method.

Human Activity Recognition (HAR) is one of the critical subjects of research in health and human machine interaction fields in recent years. Algorithms such as Support Vector Machine (SVM), K-Nearest Neighbors (K-NN), Decision Tree (DT) and many other algorithms were previously implemented to serve this common goal but most of the traditional Machine learning proposed solutions were not satisfying in term of accuracy and real time testing process. For that, a human activities analysis and recognition system with an embedded trained ANN model on Raspberry PI for an online testing process is proposed in this work. This paper includes a comparative study between the Artificial Neural Network (ANN) and the Recurrent Neural Network (RNN), using signals produced by the accelerometer and gyroscope, embedded within the BlueNRG-Tile sensor. After evaluate algorithms performance in terms of accuracy and precision which reached an accuracy of 82% for ANN and 99% for RNN, obtained ANN model was implemented in a Raspberry PI for real-time predictions. Results show that the system provides a real-time human activity recognition with an accuracy of 86%.

The omnipresent connectivity and highly distributed intelligence of Internet of Things (IoTs) integrated with Wireless Sensor Networks (WSNs) emerge as the backbone of advanced remote surveillance and control. A reliable Smart-Lab system using IoT technologies is designed to control the entire activities of a research laboratory. Smart-Lab provides a real IT intelligent environment by exploring the interaction of devices under a single roof enabling a better alignment of the research with researchers and end-users. To this end, heterogeneous IoT devices are integrated into a centralized monitoring system with developed Application Programming Interface (API). The designed API conceals the complexities of wireless communication protocols, and provides a well-defined and easy-to-use way to collect data from sensors. To implement the back-end, a Representational State Transfer (REST) approach is used. Results of different test-beds can be, stored in a MySQL database and visualized for users on the web page through the use of JSON. Then, as a proof of concept, a real-time IoT based environmental monitoring system is designed.