Brain–machine interfaces (BMI) for lower-limb exoskeletons are a state-of-the-art neurorehabilitation modality. They decode electroencephalographic (EEG) recordings during motor imagery (MI)—the mental rehearsal of movement—to infer intent and drive exoskeleton control. Yet MI decoding suffers from low signal-to-noise ratio, EEG non-stationarity, and high inter-trial/subject variability. Conventional machine-learning classifiers further struggle with limited training data and over-fitting, undermining real-time robustness. In this preliminary, offline study on a single subject, a novel semi-supervised MI-classification network is implemented that includes an L2-normalized autoencoder with dual reconstruction and classification branches—that, to our knowledge, is the first correctly tailored for closed-loop lower-limb exoskeleton control. This method is compared against four supervised approaches using a hybrid feature-extraction pipeline capturing spectral, spatial, and temporal EEG dynamics. Supervised models were evaluated via leave-one-out cross-validation, while the semi-supervised framework’s latent representations were examined with K-means clustering and t-Stochastic Neighbour Embeddings (t-SNE). Event-based false-positive (FPR) and true-positive ratios (TPR) served as comparative metrics. All approaches achieved 61–67 % accuracy, with the semi-supervised network showing a lower FPR—suggesting its promise for more robust, data-efficient BMI-driven exoskeleton control.

A new approach that includes the detection of Error Related Potentials (ErrP) for self-tuning wrong commands in MI-BMI, with the aim of improving the accuracy of a lower limb exoskeleton gait initiation system, is currently in its early stages of development. Due to the requirement of warning the subject before the command is executed, a different type of stimulus must be used to evoke the ErrP in cases where the exoskeleton is about to move against the subject's will. Therefore, it is essential to research a feedback type that better differentiates the ErrP from the correct cases, in order to achieve efficient performance of the BMI. As such, we have analyzed both Tactile (T) and VisuoTactile (VT) feedbacks to not only verify the realism of the designed protocol, but also to examine their effectiveness in eliciting ErrP.

Abstract Non-invasive brain-computer interfaces can be implemented through different paradigms, the most used one being motor imagery and evoked potentials, although recently there has been an interest in paradigms based on perception and visual imagery. Following this approach, this work demonstrates the classification of visual imagery, visual perception and also the possibility of knowledge transfer between these two domains from EEG signals using convolutional neural networks. Also, we propose an adequate framework for such classification, which uses convolutional neural networks and the black hole heuristic algorithm for the search for optimal neural network structures.

Brain-switches are systems that allow a binary communication between the brain and the outside world without using the peripheral nervous system. These systems are of interest since they can be executed with high success asynchronously and, although the amount of information processed per time unit is low, they represent a first step towards a full Brain-Computer Interface system. In this work, the possibility of creating a Brain-Switch using visual perception has been studied. Using the Black Hole search heuristic algorithm and Convolutional Neural Networks (CNN) in time domain, we obtain an average of 86% in 36 subjects with a Cohen's kappa value of 0.73.

Motor imagery (MI) is defined as the process of imaging the execution of a movement. This brain task has been used as a control paradigm for brain-computer interfaces (BCI). A BCI has the objective to decode brain patterns and translate them into commands to provide a communication with output devices. In this work, three different approaches have been compared for the design of a lower-limb MI BCI that will control the activation/deactivation of a treadmill: features from alpha, beta and lower gamma frequency band. The average accuracy for training trials was 71.43 ±10.62%. and the average accuracy for test trials was 70.40 ±11.09%.

Brain-Computer Interfaces (BCIs) are becoming an important technological tool for the rehabilitation process of patients with locomotor problems, due to their ability to recover the connection between brain and limbs by promoting neural plasticity. They can be used as assistive devices to improve the mobility of handicapped people. For this reason, current BCIs have to be improved to allow an accurate and natural use of external devices. This work proposes a novel methodology for the detection of the intention to change the direction during gait based on event-related desynchronization (ERD). Frequency and temporal features of the electroencephalographic (EEG) signals are characterized. Then, a selection of the most influential features and electrodes to differentiate the direction change intention from the walking is carried out. Best results are obtained when combining frequency and temporal features with an average accuracy of [Formula: see text]%, which are promising to be applied for future BCIs.

The use of transcranial direct current stimulation (tDCS) has been related to the improvement of motor and learning tasks. The current research studies the effects of an asymmetric tDCS setup over brain connectivity, when the subject is performing a motor imagery (MI) task during five consecutive days. A brain-computer interface (BCI) based on electroencephalography is simulated in offline analysis to study the effect that tDCS has over different electrode configurations for the BCI. This way, the BCI performance is used as a validation index of the effect of the tDCS setup by the analysis of the classifier accuracy of the experimental sessions. In addition, the relationship between the brain connectivity and the BCI accuracy performance is analyzed. Results indicate that tDCS group, in comparison to the placebo sham group, shows a higher significant number of connectivity interactions in the motor electrodes during MI tasks and an increasing BCI accuracy over the days. However, the asymmetric tDCS setup does not improve the BCI performance of the electrodes in the intended hemisphere.

The use of a brain-machine interface (BMI) in combination with powered exoskeletons can assist patients with lower limb disabilities to walk again. These neurorobotic systems are commonly based on motor imagery, but their performance may suffer from lack of user engagement in the task or from cognitive load due to multi-tasking. The present paper shows a novel algorithm based on the gamma spectral band, using the Stockwell transform and a set of smoothing filters, to assess the quality of and improve the decoding of motor imagery during the use of a BMI-Rex exoskeleton system. The results computed in a pseudo-online scenario reveal a high accuracy with a very low false positive ratio.

Electroencephalography signals can be used to detect start and stop times of gait. This is useful for people who have lost or present serial low limb motor difficulties in order to work in conjunction with an exoskeleton. Normally, the frequency bands that are used to detect the gait or stop intentions are related to mu and beta frequency bands. However, in order to enhance the electroencephalography signal quality, it is necessary to increase the signal-to-noise ratio. In the paper, a former research is complemented with the use of different types of frequency and spatial filters. A multi resolution analysis tool based on Hilbert-Huang transform is also introduced as a new processing tool and its results discussed with the help of a recent developed comparison index.

Restoring the gait cycle is vital in motor-impaired people. To accomplish this, it is necessary to study the brain signals in different areas. This work analyzes EEG data offline and pseudo-online for different electrode configurations and different processing-time windows to detect the pedaling start initiation. Premotor cortex is the area related to movement intention. Therefore, in this study, the FZ electrode, which is located in this area, was included in the analysis of the electrode configurations, testing whether it plays an important role in the detection of pedaling start intention. Results show that using time before and after the movement onset for processing is preferred. The presence of the FZ electrode seems to be desirable when analyzing data offline, but is not statistically significant when analyzing data pseudo-online. This suggests the FZ electrode could be ignored when analyzing data in real time, since the processing is the same as pseudo-online.

In recent years, several studies have shown that there is a correlation between electroencephalographic (EEG) signals and hand-reaching kinematic parameters after applying linear decoders. These studies have been generally conducted using actual upper limb movements, but so far there has been little discussion about the possibility of applying these decoders to motor imagery tasks. Moreover, the use of these decoders is rather controversial and there is no general agreement about the metrics used to compare decoded and real kinematics. In this paper, we have applied this methodology to upper limb imagined movements using a center-out protocol. Our results show that, although decoding performance is poor, there are significant components, particularly in horizontal imagined movements, that could be translated into reliable output commands. For this purpose, we have proposed a discrete classification of reached targets showing significant classification rates when the number of classified targets decreases.

Walking is for humans an essential task in our daily life. However, there is a huge (and growing) number of people who have this ability diminished or are not able to walk due to motor disabilities. In this paper, a system to detect the start and the stop of the gait through electroencephalographic signals has been developed. The system has been designed in order to be applied in the future to control a lower limb exoskeleton to help stroke or spinal cord injured patients during the gait. The brain-machine interface (BMI) training has been optimized through a preliminary analysis using the brain information recorded during the experiments performed by three healthy subjects. Afterward, the system has been verified by other four healthy subjects and three patients in a real-time test. In both preliminary optimization analysis and real-time tests, the results obtained are very similar. The true positive rates are [Formula: see text] and [Formula: see text] respectively. Regarding the false positive per minute, the values are also very similar, decreasing from 2.66 in preliminary tests to 1.90 in real-time. Finally, the average latencies in the detection of the movement intentions are 794 and 798[Formula: see text]ms, preliminary and real-time tests respectively.

Regain the ability of walking represents a great progress for people with disabilities. The main goal of this paper is the verification of a method to detect the intention of starting and finishing the gait. This method was checked using recordings from spinal cord injury patients. The system has been designed to be part of the control of a lower limb wearable exoskeleton which will be used not only in the rehabilitation process but also for assistive tasks. Four patients took part in this experiment. Three of them obtained hopeful results achieving a high rate in the detection of the start and stop intentions (68.6% in averaged) with a low rate of wrong classification (around 1.51 wrong detection per minute in average), obtaining a good accuracy of the system (around 80.0%). However, the system does not seem accurate in one of the patients (P3).

In this paper, linear regression models will be used to decode individual joint angles from low frequency EEG components. To that end, isotonic flexion/extension knee movements will be analyzed. Particularly, the decoding performance of healthy and incomplete spinal cord injured subjects will be assessed to determine the behavior of this methodology with motor disabled people. When studying cortical activity during walking, the appearance of muscular artifacts severely influences the EEG signals recorded. The analysis of single joint movements should decrease the noise provoked by the gait process itself. Additionally, different time windows prior to the decoded angle will be assessed to obtain a more reliable decoder. The results show that decoding performance is significantly above chance for most of the subjects (both healthy and disabled) and suggests that meaningful information of the movement planning starts around 2.5 seconds prior to the decoded angle.

On this work several studies related to the analysis of electroencephalographically signals to evaluate the cognitive mechanisms during human gait are compiled. This research is performed in the framework of an European project (BioMot project) oriented to the development of a lower limb exoskeleton controlled by signals from physiological and kinematic sources. The project includes the study of the cognitive mechanisms and volitional intentions of the exoskeleton wearer as well as other physical information provided from other sources. This paper is oriented on those studies related with the users' cognitive mechanisms during gait. The phenomena evaluated are the changes in the brain waves depending on the attention that a user is paying in the gait and the event-related potential evoked when an obstacle suddenly appear during walking. Initial experiments with healthy users provide encouraging results, obtaining an average value of 65,25% in the classification of four levels of attention and a potential related with the alertness state experienced by a user when an obstacle appears is characterized.

In this work, a study that analyzes the best combinations of mental tasks in a Brain-Computer Interface (BCI) using a classifier based on Support Vector Machine (SVM) is presented. To that end, 12 mental tasks of different nature are analyzed and the results of the classification for the combinations of two, three and four tasks are obtained. Four volunteers performed registers of the 12 tasks. The main goal is to find the combination of more than three mental tasks that obtains the highest reliability to apply it in future complex applications that require the use of more than three control commands. After a selection procedure, the results obtained show higher success rates. Using the information provided by every single electrode, an average of 87.10% is obtained as success rate for the classification of two mental tasks, 65.67% for three mental tasks and 50.76% for four mental tasks. Moreover combinations of the best electrodes are studied, improving the accuracy of the system. Using the best five electrodes, averages of 91.42%, 72.89% and 59.75% are obtained classifying two, three and four mental tasks respectively. These results suggest that it is possible to differentiate with enough reliability between more than three mental tasks using the methodology proposed.

Recent studies have hypothesized that the motor cortex is particularly active during specific phases of gait cycle. It has been found that cortical coherence appearance differs in time depending on walking speed. In this work, we analyze the influence of walking speed by decoding knee angles from low frequency EEG components. Linear regression models are applied to show significant correlations between actual and decoded angles while different walking speeds are performed. Additionally, a comparison between walking speeds suggests that the decoding correlation increases with lower speeds.

This letter describes a man-machine bimodal interface that is based on the combination of visual control, through electrooculography, and manual control, using a desktop input device that is operated with the hand. This letter introduces design concepts for this bimodal interface with the goal to support those that cannot use desktop device controls, those that require support with desktop device controls, as well as those with normative capabilities but who wish to enhance speed performance. To that end, several control strategies have been designed. Each control strategy has been tested in applications with a real robot arm. The results indicate that such bimodal interfaces show promise to support a range of motor limitations.

In this paper, two control strategies to move a planar robot arm with a non-invasive spontaneous brain-machine interface (BMI) have been compared. The BMI is based on the correlation of EEG maps and allows differentiating between two mental tasks related to motor imagery. Using the BMI, the user is able to control 2D movements of the robot arm in order to reach several goals. The first control strategy is based on a hierarchical control and the second one uses a directional control of the movement. The robot arm used is the PuParm, a force-controlled planar robot designed and developed by the nBio research group at the Miguel Hernández University of Elche (Spain). Three goals have been placed on the experimental setup. After performing the tests, time taken to reach the goals and errors have been presented and compared, showing the advantages and disadvantages of each strategy. The evidence from this study suggests that the control of a planar robot is possible with both strategies. The hierarchical control is slower but more reliable, while the directional control is much faster and more relaxing for the user, but less precise. These findings indicate that future assistive applications like grasping daily objects in a realistic environment could be performed with this system.

This paper describes a Brain-Computer Interface (BCI) based on electroencephalography (EEG). This BCI registers the brain rhythmic activity through 16 electrodes situated on the scalp and differentiates three cognitive processes. The Wavelet Transform (WT) has been used to extract the features. In order to differentiate three mental tasks, two Linear Discriminant Analysis (LDA) based classifiers have been developed. Both classifiers integrate four simultaneous LDA model solutions. One classifier uses a score system to classify the EEG features vector, while the other classifier applies a logical criterion. In the paper, both classifiers have been evaluated. The experimental results with different volunteers have been reported in the paper. This BCI will be incorporated into a shared control architecture that we are developing to control a robot arm. This shared control architecture is based on Radio Frequency Identification (RFID) technology.

This paper describes a multimodal interface that combines ocular information, through electrooculography, and haptics information, with a desktop-based haptics feedback device. Two control strategies are defined in order to test the advantages of multimodal interaction with an external device, in this case a robot arm from Fanuc. Three applications have been designed using the previously described control strategies. These applications are aimed at studying the usefulness of this kind of interfaces as a substitute of classical man-machine devices.

Using haptic (the sensing of touch) technology as an interface in medical and surgical procedures is a large interesting goal because of the benefits involved. This work presents a developed tool for evaluating the performance of a classic 2D-3D processing of a stack of medical preoperative images, and a new version of an efficient and simple algorithm for the integration of the haptic sense in a medical 3D environment generated from the 3D reconstruction. First of all, the developed tool for the 2D segmentation and 3D reconstruction is presented. The classical pipeline for surface 3D reconstruction is reviewed under a parametrical point of view. These parameters will play an important role in the analysis of the haptic behavior. Besides, all the parameters of the reconstruction are accessible and can be modified on-line during the reconstruction procedure. Later, the software architecture used for the integration of the haptic devices is described. The haptic rendering algorithm is detailed, including the collision detection algorithm (a simple ray-tracing scheme programmed using VTK capacities) that is used with the medical images. Finally, some results of the evaluation of the behavior of this algorithm are resumed.

This paper describes a new methodology for designing bilateral controllers based on transparency that applies a modified scheme of control by state of convergence. The design is based on modelling the behavior of the master and slave which regard state space equations, and also taking into account that perfect transparency cannot be reached. This methodology allows designing the controllers in order to obtain the convergence between the state of the master and the slave. Furthermore, it consequently provides a higher degree of transparency to the operator. The paper explains criteria in achieving convergence between the master and the slave, and so as with transparency on a steady state. A set of equations that calculate controller gains have been obtained by applying such criteria. In order to verify this new methodology, a master-slave system of 3 DoF have been used.

This paper presents a new bilateral control scheme for time-delayed teleoperation designed to achieve transparency. The control scheme allows that the slave follows the master in spite of the time delay, ant that the force displayed to the operator was exactly the reaction force from the environment. In addition, the interaction force of the slave with the environment is adapted to the master/slave ratio when it is reflected to the operator, improving the transparency of the system. The bilateral control scheme can be used in contact situations or non-contact situations of the slave with the environment. Together with the control scheme, the paper describes an analytical design method that allows the obtaining of the control gains.

This paper presents a discrete design and control method of teleoperation systems. The design method is based on the state space formulation and it allows to obtain the control gains for any teleoperation system where the master and the slave manipulators would be represented by nth-order discrete linear transfer functions. The control method allows that the slave follows the master through the state convergence between the master and the slave. The method is able also to establish the desire dynamics of this convergence and the dynamics of the slave manipulator. The advantage of this method is that it can be applied to control a teleoperation system where the master and the slave are modeled by transfer functions of different order. In order to validate the control method, simulation results are shown considering a teleoperation system where the master and the slave are modeled by discrete transfer functions of different order