Alberto López-Delis

Adaptive Spatial Filter Based on Similarity Indices to Preserve the Neural Information on EEG Signals during On-Line Processing

This work presents a new on-line adaptive filter, which is based on a similarity analysis between standard electrode locations, in order to reduce artifacts and common interferences throughout electroencephalography (EEG) signals, but preserving the useful information. Standard deviation and Concordance Correlation Coefficient (CCC) between target electrodes and its correspondent neighbor electrodes are analyzed on sliding windows to select those neighbors that are highly correlated. Afterwards, a model based on CCC is applied to provide higher values of weight to those correlated electrodes with lower similarity to the target electrode. The approach was applied to brain computer-interfaces (BCIs) based on Canonical Correlation Analysis (CCA) to recognize 40 targets of steady-state visual evoked potential (SSVEP), providing an accuracy (ACC) of 86.44 ± 2.81%. In addition, also using this approach, features of low frequency were selected in the pre-processing stage of another BCI to recognize gait planning. In this case, the recognition was significantly (p<0.01) improved for most of the subjects (ACC≥74.79%), when compared with other BCIs based on Common Spatial Pattern, Filter Bank-Common Spatial Pattern, and Riemannian Geometry.

Onset and Peak Detection over Pulse Wave Using Supervised SOM Network

temporal windowing, which is presented to the network to decide whether the window corresponds to an onset or peak in the pulse wave. Results of the classification reach 97.93% over the validation dataset. Sensitivity and positive predictivity measures were used to assess the proposed method, reaching 100% for sensitivity and 99.84% for the positive predictivity detecting peaks in the signals. This proposal takes advantages from SOM neural networks for pattern classification and detection. Additionally, ECG signal is not necessary in the presented methodology.

Onset and Peak Pattern Recognition on Photoplethysmographic Signals Using Neural Networks

Traditional methodologies use electrocardiographic ECG signals to develop automatic methods for onset and peak detection on the arterial pulse wave. In the present work a Multilayer Perceptron MLP neural network is used for classifying fiducial points on photoplethysmographic PPG signals. System was trained with a dataset of temporal segments from signals located based on information about onset and peak points. Different segments sizes and units in the neural network were used for the classification, and optimal values were searched. Results of the classification reach 98.1% in worse of cases. This proposal takes advantages from MLP neural networks for pattern classification. Additionally, the use of ECG signal was avoided in the presented methodology, making the system robust, less expensive and portable in front of this problem.

A Comparison of Myoelectric Pattern Recognition Methods to Control an Upper Limb Active Exoskeleton

Physically impaired people may use Surface Electromyography (sEMG) signals to control assistive devices in an automatic way. sEMG signals directly reflect the human motion intention, they can be used as input information for active exoskeleton control. This paper proposes a set of myoelectric algorithms based on machine learning for detecting movement intention aimed at controlling an upper limb active exoskeleton. The algorithms use a feature extraction stage based on a combination of time and frequency domain features (mean absolute value – waveform length, and auto-regressive model, respectively). The pattern recognition stage uses Linear Discriminant Analysis, K-Nearest Neighbor, Support Vector Machine and Bayesian classifiers. Additionally, two post-processing techniques are incorporated: majority vote and transition removal. The performance of the algorithms is evaluated with parameters of sensitivity, specificity, positive predictive value, error rate and active error rate, under typical conditions. These evaluations allow identifying pattern recognition algorithms for real-time control of an active exoskeleton.Physically impaired people may use Surface Electromyography (sEMG) signals to control assistive devices in an automatic way. sEMG signals directly reflect the human motion intention, they can be used as input information for active exoskeleton control. This paper proposes a set of myoelectric algorithms based on machine learning for detecting movement intention aimed at controlling an upper limb active exoskeleton. The algorithms use a feature extraction stage based on a combination of time and frequency domain features (mean absolute value – waveform length, and auto-regressive model, respectively). The pattern recognition stage uses Linear Discriminant Analysis, K-Nearest Neighbor, Support Vector Machine and Bayesian classifiers. Additionally, two post-processing techniques are incorporated: majority vote and transition removal. The performance of the algorithms is evaluated with parameters of sensitivity, specificity, positive predictive value, error rate and active error rate, under typical conditions. These evaluations allow identifying pattern recognition algorithms for real-time control of an active exoskeleton.

Using linear discriminant function to detect eyes closing activities through alpha wave

This work presents an alternative method to detect events correlated to eyes opening and closing, based on electroencephalography (EEG) measured from the occipital lobe. The goal is to propose a method based on linear discriminant function to classify segments of EEG signals that contain activities originated by eyes closing. A linear discriminant function presented by Fisher is employed to detect these activities on segments of 2s. This method showed a good values of sensitivity (SE ≥ 85 %) and specificity (SP ≥ 60 %). This approach can be used to control the switching of a brain computer interface (BCI).

Surface EMG signal analysis based on the empirical mode decomposition for human-robot interaction

Surface Electromyography (SEMG) is the electrical manifestation of the neuromuscular activation associated with a contracting muscle. SEMG directly reflects the human motion intention; thus, they can be used as input information for human-robot interaction. Taking into account that SEMG signals are complex physiological signals, being nonlinear, non-stationary and non-periodic, myoelectric classification methods must take into account such characteristics to be more effective. Recently, a novel technique for analysis of nonlinear and non-stationary signals was successfully applied to several kinds of investigations including seismological and biological signals. This technique, named Hilbert-Huang Transform (HHT) is formed by two complementary tools, which are called empirical mode decomposition (EMD) and Hilbert spectrum (HS). This work proposes a novel EMD-based myoelectric pattern recognition technique to be applied in human-robot interaction. The process of feature extraction is performed by two steps, firstly, the EMD decomposes the input SEMG signal into a set of functions designated as Intrinsic Mode Function (IMF); and secondly, it is calculated for each resulting IMF the RMS (Root Mean Square) and the coefficients of a four-order autoregressive model. The process of classification based on a linear classifier (Linear Discriminant Analysis). Using a database of EMG signals, the proposed method was applied to classify human upper-limb motion via EMG signals. The database includes 8 recorded SEMG channels from forearm in the execution of 7 movements. The error of classification was 3.3%. Obtained results suggest that the proposed myoelectric pattern recognition technique may be applied in Human-Robot Interaction (HRI) to control external systems such an upper limb motor neuroprosthesis.

A Real-Time Research Platform for Intent Pattern Recognition: Implementation, Validation and Application

Despite multiple advances with myoelectric control, currently there is still an important need to develop more effective methods for controlling prosthesis and exoskeletons in a natural way. This work describes the design and development of a research tool for the design, development and evaluation of algorithms of myoelectric control which base on intention detection from neuromuscular activation patterns. This platform provides integrated hardware and software tools for real-time acquisition, preprocessing, visualization, storage and analysis of biological signals. It is composed of a bio-instrumentation system controlled by a real-time software created in Simulink and executed on the xPC-target platform and, a Java based software application that allows to manage the acquisition and storage processes by a system operator. System evaluation was performed by the comparison with reference signals provided by a function generator and, as an example of the application of the developed acquisition platform, it was carried out a set of experiments to decode movements at the upper-limb level.

Non-supervised Feature Selection: Evaluation in a BCI for Single-Trial Recognition of Gait Preparation/Stop

Is presented a non-supervised method for feature selection based on similarity index, which is applied in a brain-computer interface (BCI) to recognize gait preparation/stops. Maximal information compression index is here used to obtain redundancies, while representation entropy value is employed to find the feature vectors with high entropy. EEG signals of six subjects were acquired on the primary cortex during walking, in order to evaluate this approach in a BCI. The maximum accuracy was 55 % and 85 % to recognize gait preparation/stops, respectively. Thus, this method can be used in a BCI to improve the time delay during dimensionality reduction.

Toward an Upper-Limb Neurorehabilitation Platform Based on FES-Assisted Bilateral Movement: Decoding User’s Intentionality

In the last years there has been a noticeable progress in motor learning, neuroplasticity and functional recovery after the occurrence of brain lesion. Rehabilitation of motor function has been associated to motor learning that occurs during repetitive, frequent and intensive training. Neuro-rehabilitation is based on the assumption that motor learning principles can be applied to motor recovery after injury, and that training can lead to permanent improvements of motor functions in patients with muscle deficits. The emergent research field of Rehabilitation Engineering may provide promise technologies for neuro-rehabilitation therapies, exploiting the motor learning and neural plasticity concepts. Among those technologies, the FES-assisted systems could provide repetitive training-based therapies and have been developed to aid or control the upper and lower limbs movements in response to user’s intentionality. Surface electromyography (SEMG) reflects directly the human motion intention, so it can be used as input information to control an active FES-assisted system. The present work describes a neurorehabilitation platform at the upper-limb level, based on bilateral coordination training (i.e. mirror movements with the unaffected arm) using a close-loop active FES system controlled by user. In this way, this work presents a novel myoelectric controller for decoding movements of user to be employed in a neurorehabilitation platform. It was carried out a set of experiments to validate the myoelectric controller in classification of seven human upper-limb movements, obtaining an average classification error of 4.3%. The results suggest that the proposed myoelectric pattern recognition method may be applied to control close-loop FES system.

Knee motion pattern classification from trunk muscle based on sEMG signals.

A prominent change is being carried out in the fields of rehabilitation and assistive exoskeletons in order to actively aid or restore legged locomotion for individuals suffering from muscular impairments, muscle weakness, neurologic injury, or disabilities that affect the lower limbs. This paper presents a characterization of knee motion patterns from Surface Electromyography (sEMG) signals, measured in the Erector spinae (ES) muscle. Feature extraction (mean absolute value, waveform length and auto-regressive model) and pattern classification methods (Linear Discrimination Analysis, K-Nearest Neighborhood and Support Vector Machine) are applied for recognition of eight-movement classes. Additionally, several channels setup are analyzed to obtain a suitable electrodes array. The results were evaluated based on signals measured from lower limb using quantitative metric such as error rate, sensitivity, specificity and predictive positive value. A high accuracy (> 95%) was obtained, which suggest that it is possible to detect the knee motion intention from ES muscle, as well as to reduce the electrode number (from 2 to 3 channels) to obtain an optimal electrodes array. This implementation can be applied for myoelectric control of lower limb active exoskeletons.