This paper presents a novel method for feature extraction from electroencephalogram (EEG) signals using empirical mode decomposition (EMD). Its use is motivated by the fact that the EMD gives an effective time-frequency analysis of nonstationary signals. The intrinsic mode functions (IMF) obtained as a result of EMD give the decomposition of a signal according to its frequency components. We present the usage of upto third order temporal moments, and spectral features including spectral centroid, coefficient of variation and the spectral skew of the IMFs for feature extraction from EEG signals. These features are physiologically relevant given that the normal EEG signals have different temporal and spectral centroids, dispersions and symmetries when compared with the pathological EEG signals. The calculated features are fed into the standard support vector machine (SVM) for classification purposes. The performance of the proposed method is studied on a publicly available dataset which is designed to handle various classification problems including the identification of epilepsy patients and detection of seizures. Experiments show that good classification results are obtained using the proposed methodology for the classification of EEG signals. Our proposed method also compares favorably to other state-of-the-art feature extraction methods.

Brain-computer interfaces (BCIs) have the potential to improve functionality in chronic stoke patients when applied over a large number of sessions. Here we evaluated the effect and the underlying mechanisms of three BCI training sessions in a double-blind sham-controlled design. The applied BCI is based on Hebbian principles of associativity that hypothesize that neural assemblies activated in a correlated manner will strengthen synaptic connections. Twenty-two chronic stroke patients were divided into two training groups. Movement-related cortical potentials (MRCPs) were detected by electroencephalography during repetitions of foot dorsiflexion. Detection triggered a single electrical stimulation of the common peroneal nerve timed so that the resulting afferent volley arrived at the peak negative phase of the MRCP (BCI_associativegroup) or randomly (BCI_{nonassociative}group). Fugl-Meyer motor assessment (FM), 10-m walking speed, foot and hand tapping frequency, diffusion tensor imaging (DTI) data, and the excitability of the corticospinal tract to the target muscle [tibialis anterior (TA)] were quantified. The TA motor evoked potential (MEP) increased significantly after the BCI_associativeintervention, but not for the BCI_{nonassociative}group. FM scores (0.8 ± 0.46 point difference, P = 0.01), foot (but not finger) tapping frequency, and 10-m walking speed improved significantly for the BCI_associativegroup, indicating clinically relevant improvements. Corticospinal tract integrity on DTI did not correlate with clinical or physiological changes. For the BCI as applied here, the precise coupling between the brain command and the afferent signal was imperative for the behavioral, clinical, and neurophysiological changes reported. This association may become the driving principle for the design of BCI rehabilitation in the future. Indeed, no available BCIs can match this degree of functional improvement with such a short intervention.

Detection of movement intention from neural signals combined with assistive technologies may be used for effective neurofeedback in rehabilitation. In order to promote plasticity, a causal relation between intended actions (detected for example from the EEG) and the corresponding feedback should be established. This requires reliable detection of motor intentions. In this study, we propose a method to detect movements from EEG with limited latency. In a self-paced asynchronous BCI paradigm, the initial negative phase of the movement-related cortical potentials (MRCPs), extracted from multi-channel scalp EEG was used to detect motor execution/imagination in healthy subjects and stroke patients. For MRCP detection, it was demonstrated that a new optimized spatial filtering technique led to better accuracy than a large Laplacian spatial filter and common spatial pattern. With the optimized spatial filter, the true positive rate (TPR) for detection of movement execution in healthy subjects (n = 15) was 82.5 ± 7.8%, with latency of -66.6 ± 121 ms. Although TPR decreased with motor imagination in healthy subject (n = 10, 64.5 ± 5.33%) and with attempted movements in stroke patients (n = 5, 55.01 ± 12.01%), the results are promising for the application of this approach to provide patient-driven real-time neurofeedback.

Pattern recognition of electromyography (EMG) signals can potentially improve the performance of myoelectric control for upper limb prostheses with respect to current clinical approaches based on direct control. However, the choice of features for classification is challenging and impacts long-term performance. Here, we propose the use of EMG raw signals as direct inputs to deep networks with intrinsic feature extraction capabilities recorded over multiple days. Seven able-bodied subjects performed six active motions (plus rest), and EMG signals were recorded for 15 consecutive days with two sessions per day using the MYO armband (MYB, a wearable EMG sensor). The classification was performed by a convolutional neural network (CNN) with raw bipolar EMG samples as the inputs, and the performance was compared with linear discriminant analysis (LDA) and stacked sparse autoencoders with features (SSAE-f) and raw samples (SSAE-r) as inputs. CNN outperformed (lower classification error) both LDA and SSAE-r in the within-session, between sessions on same day, between the pair of days, and leave-out one-day evaluation (p < 0.001) analyses. However, no significant difference was found between CNN and SSAE-f. These results demonstrated that CNN significantly improved performance and increased robustness over time compared with standard LDA with associated handcrafted features. This data-driven features extraction approach may overcome the problem of the feature calibration and selection in myoelectric control.

Gait is the locomotion attained through the movement of limbs and gait analysis examines the patterns (normal/abnormal) depending on the gait cycle. It contributes to the development of various applications in the medical, security, sports, and fitness domains to improve the overall outcome. Among many available technologies, two emerging technologies that play a central role in modern day gait analysis are: A) wearable sensors which provide a convenient, efficient, and inexpensive way to collect data and B) Machine Learning Methods (MLMs) which enable high accuracy gait feature extraction for analysis. Given their prominent roles, this paper presents a review of the latest trends in gait analysis using wearable sensors and Machine Learning (ML). It explores the recent papers along with the publication details and key parameters such as sampling rates, MLMs, wearable sensors, number of sensors, and their locations. Furthermore, the paper provides recommendations for selecting a MLM, wearable sensor and its location for a specific application. Finally, it suggests some future directions for gait analysis and its applications.