Daniel Planelles

SVM-based Brain–Machine Interface for controlling a robot arm through four mental tasks

Abstract Human–Machine Interfaces can be very useful to improve the quality of life of physically impaired users. In this work, a non-invasive spontaneous Brain–Machine Interface (BMI) has been designed to control a robot arm through the mental activity of the users. This BMI uses the classification of four mental tasks in order to manage the movements of the robot. The high accuracy in the classification of these tasks (around 70%) allows a quick accomplishment of the experiment designed, even with the low signal-to-noise ratio of this kind of signals. The experiment consists of reaching four points in the workspace of an industrial robot in the established order. After a brief training, the volunteers are able to control the robot in a real time activity. The real time test shows that the system can be applied to do more complex activity such as pick and place tasks if a supplementary system is added. These interfaces are very adequate in the control of rehabilitation or assistance systems for people suffering from motor impairment.

Using a brain-machine interface to control a hybrid upper limb exoskeleton during rehabilitation of patients with neurological conditions

BackgroundAs a consequence of the increase of cerebro-vascular accidents, the number of people suffering from motor disabilities is raising. Exoskeletons, Functional Electrical Stimulation (FES) devices and Brain-Machine Interfaces (BMIs) could be combined for rehabilitation purposes in order to improve therapy outcomes.MethodsIn this work, a system based on a hybrid upper limb exoskeleton is used for neurological rehabilitation. Reaching movements are supported by the passive exoskeleton ArmeoSpring and FES. The movement execution is triggered by an EEG-based BMI. The BMI uses two different methods to interact with the exoskeleton from the user’s brain activity. The first method relies on motor imagery tasks classification, whilst the second one is based on movement intention detection.ResultsThree healthy users and five patients with neurological conditions participated in the experiments to verify the usability of the system. Using the BMI based on motor imagery, healthy volunteers obtained an average accuracy of 82.9 ± 14.5 %, and patients obtained an accuracy of 65.3 ± 9.0 %, with a low False Positives rate (FP) (19.2 ± 10.4 % and 15.0 ± 8.4 %, respectively). On the other hand, by using the BMI based on detecting the arm movement intention, the average accuracy was 76.7 ± 13.2 % for healthy users and 71.6 ± 15.8 % for patients, with 28.7 ± 19.9 % and 21.2 ± 13.3 % of FP rate (healthy users and patients, respectively).ConclusionsThe accuracy of the results shows that the combined use of a hybrid upper limb exoskeleton and a BMI could be used for rehabilitation therapies. The advantage of this system is that the user is an active part of the rehabilitation procedure. The next step will be to verify what are the clinical benefits for the patients using this new rehabilitation procedure.

Evaluating Classifiers to Detect Arm Movement Intention from EEG Signals

This paper presents a methodology to detect the intention to make a reaching movement with the arm in healthy subjects before the movement actually starts. This is done by measuring brain activity through electroencephalographic (EEG) signals that are registered by electrodes placed over the scalp. The preparation and performance of an arm movement generate a phenomenon called event-related desynchronization (ERD) in the mu and beta frequency bands. A novel methodology to characterize this cognitive process based on three sums of power spectral frequencies involved in ERD is presented. The main objective of this paper is to set the benchmark for classifiers and to choose the most convenient. The best results are obtained using an SVM classifier with around 72% accuracy. This classifier will be used in further research to generate the control commands to move a robotic exoskeleton that helps people suffering from motor disabilities to perform the movement. The final aim is that this brain-controlled robotic exoskeleton improves the current rehabilitation processes of disabled people.

Online classification of two mental tasks using a SVM-based BCI system

A Brain-Computer Interface (BCI) can be very useful to help people with several movement disabilities to improve their independence or to assist them in rehabilitation tasks. In this paper, the results of the online classification of two mental tasks from electroencephalographic signals (EEG) are shown. The objective of this paper is to determine whether the accuracy in the online differentiation of two mental tasks could be enough to command a robot arm using two mental tasks. The results demonstrate that the features obtained using periodogram (a Power Spectral Density estimation) and the classification of these using a SVM-based (Support Vector Machine) system can be used in a reliable control of a robot arm. For all the users, the accuracy is around 87±2%. This accuracy is enough to be used to this end in real time.

First steps in the development of an EEG-based system to detect intention of gait initiation

The ability to walk is a very important characteristic of the human being that unfortunately not everyone can enjoy. Subjects with spinal cord damage or who have had a stroke may have difficulties in walking, or even be unable to walk, depending on their degree of disability. Although, in some cases, it is possible to recover mobility, the rehabilitation process can continue for long periods of time. This article presents a methodology to detect walking intention in healthy subjects before the movement actually starts. This is done by measuring brain activity through electroencephalographic signals (EEG) that can be extracted from the motor cortex. The preparation and performance of a movement generates a phenomenon called event-related desynchronization (ERD) in the mu and beta frequency bands. This paper shows a technique to characterize this brain process during gait onset and the results of preliminary tests with 3 healthy subjects. The output of the classifier can serve as a command to move an exoskeletal system that helps to perform the movement, so it might improve rehabilitation. The aim is to obtain reliable results to continue further research with more healthy subjects and patients who are not able to walk or do not have the strength to do so and also carry out real time tests.

Detection of Gait Initiation Through a ERD-Based Brain-Computer Interface

In this paper, an experiment designed to detect the will to perform several steps forward (as gait initiation) before it occurs using the electroencephalographic (EEG) signals collected from the scalp is presented. In order to detect this movement intention, the Event-Related Desynchronization phenomenon is detected using a SVM-based classifier. The preliminary results from seven users have been presented. In this work, the results obtained in a previous paper are enhance obtaining similar true positive rates (around 66 % in average) but reducing the false positive rates (with an average around 20 %). In the future, this improved Brain-Computer Interface will be used as part of the control system of an exoskeleton attached to the lower limb of people with incomplete and complete spinal cord injury to initiate their gait cycle.

Passive robot assistance in arm movement decoding from EEG signals

The past decade has seen the rapid development of upper limb kinematic decoding techniques by performing intracortical recordings of brain signals. However, the use of non-invasive approaches to perform similar decoding procedures is still in its early stages. Previous works show that there is a correlation between electroencephalographic signals and position/velocity hand movement parameters. In these studies, it was observed that uniform and linear movements were better decoded. In this paper, a passive robot assistance has been used during non-invasive hand kinematics decoding. To study the influence of movement variability in this decoding, we propose to use this passive robot arm to track a circle that moves on the screen. Brain signals are recorded while performing the movements. The decoding method is based on a linear regression model applied to low frequency signal components. The decoding accuracy has been analyzed by studying the correlation between the reconstructed and original movements. These results have been compared to previous works in which a computer mouse was used to perform the movements. The results show more stable and realistic decoding accuracies when using the planar robot. Also, we show that decoding accuracy is directly correlated to tracking success.

Selection of the best mental tasks for a SVM-based BCI system.

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.

Brain-Machine Interface system to differentiate between five mental tasks

The large amount of patients suffering from motor disabilities has motivated a lot of studies in order to improve their mobility and quality of life. A Brain-Machine Interface (BMI) can be very useful to control a system that is able to improve the independence of people with motor disabilities. The electroencephalographic (EEG) signals are commonly used to control systems as a robot arm or other devices like rehabilitation systems. Motor imagery is one of the techniques that is usually used to command a BMI. To that end, an accurate method to classify different mental tasks is needed. In this paper, the accuracy of a SVM-based system (Support Vector Machine) is analyzed using four different procedures that include two feature extraction methods: Periodogram and Welchs method. The results show that by using a SVM-based system it is possible to obtain enough accuracy for the suggested purpose. The system defined in this work is able to distinguish between five different mental tasks with a considerably higher accuracy than the random behavior (20% for five tasks). The average success rate for three users is 47,75±4%. Using five different tasks, it is possible to control the movement of a robotic arm in a 2-D plane, assigning a task for each direction (left, right, forward and backward) and another for a rest state.

Processing EEG Signals to Detect Intention of Upper Limb Movement

In the world there is a large number of people who have trouble performing movements that are simple for others, such as people who have suffered a stroke or have damage in the spinal cord. However, thanks to neuroscience, there is knowledge about the cognitive processes that occur in the brain and it is possible to help these people by using brain-computer interfaces. In this paper the movement of the arm of a healthy person is under research. Different processing methods and classifiers are studied in order to obtain the minimum false positive rate with the best true positive rate to detect the intention to make such movement in future real time tests. The ultimate goal is to use this system with an exoskeleton attached to the user arm. Thus, these kind of disabled people will perform movements on their own will by activating the exoskeleton joints. This system could be used in motor rehabilitation processes since it would allow the patient to recover the damaged communication channels of the brain or create new ones due to brain plasticity.