Hongyi Li

A hybrid BCI based on EEG and fNIRS signals improves the performance of decoding motor imagery of both force and speed of hand clenching

OBJECTIVEnIn order to increase the number of states classified by a brain-computer interface (BCI), we utilized a motor imagery task where subjects imagined both force and speed of hand clenching.nnnAPPROACHnThe BCI utilized simultaneously recorded electroencephalographic (EEG) and functional near-infrared spectroscopy (fNIRS) signals. The time-phase-frequency feature was extracted from EEG, whereas the HbD [the difference of oxy-hemoglobin (HbO) and deoxy-hemoglobin (Hb)] feature was used to improve the classification accuracy of fNIRS. The EEG and fNIRS features were combined and optimized using the joint mutual information (JMI) feature selection criterion; then the extracted features were classified with the extreme learning machines (ELMs).nnnMAIN RESULTSnIn this study, the averaged classification accuracy of EEG signals achieved by the time-phase-frequency feature improved by 7%, to 18%, more than the single-type feature, and improved by 15% more than common spatial pattern (CSP) feature. The HbD feature of fNIRS signals improved the accuracy by 1%, to 4%, more than Hb, HbO, or HbT (total hemoglobin). The EEG-fNIRS feature for decoding motor imagery of both force and speed of hand clenching achieved an accuracy of 89%xa0±xa02%, and improved the accuracy by 1% to 5% more than the sole EEG or fNIRS feature.nnnSIGNIFICANCEnOur novel motor imagery paradigm improves BCI performance by increasing the number of extracted commands. Both the time-phase-frequency and the HbD feature improve the classification accuracy of EEG and fNIRS signals, respectively, and the hybrid EEG-fNIRS technique achieves a higher decoding accuracy for two-class motor imagery, which may provide the framework for future multi-modal online BCI systems.

Classification of Hemodynamic Responses Associated With Force and Speed Imagery for a Brain-Computer Interface

Functional near-infrared spectroscopy (fNIRS) is an emerging optical technique, which can assess brain activities associated with tasks. In this study, six participants were asked to perform three imageries of hand clenching associated with force and speed, respectively. Joint mutual information (JMI) criterion was used to extract the optimal features of hemodynamic responses. And extreme learning machine (ELM) was employed to be the classifier. ELM solved the major bottleneck of feedforward neural networks in learning speed, this classifier was easily implemented and less sensitive to specified parameters. The 2-class fNIRS-BCI system was firstly built with an average accuracy of 76.7xa0%, when all force and speed tasks were categorized as one class, respectively. The multi-class systems based on different levels of force and speed attempted to be investigated, the accuracies were moderate. This study provided a novel paradigm for establishing fNIRS-BCI system, and provided a possibility to produce more degrees of freedom in BCI system.

NIRS-based classification of clench force and speed motor imagery with the use of empirical mode decomposition for BCI.

Near-infrared spectroscopy (NIRS) is a non-invasive optical technique used for brain-computer interface (BCI). This study aims to investigate the brain hemodynamic responses of clench force and speed motor imagery and extract task-relevant features to obtain better classification performance. Given the non-stationary characteristics of real hemodynamic measurements, empirical mode decomposition (EMD) was applied to reduce the physiological noise overwhelmed in the task-relevant NIRS signals. Compared with continuous wavelet decomposition, EMD does not require a pre-determined basis function. EMD decomposes the original signals into a set of intrinsic mode functions (IMFs). In this study, joint mutual information was applied to select the optimal features, and support vector machine was used as a classifier. Offline and pseudo-online analyses showed that the most feasible classification accuracy can be obtained using IMFs as input features. Accordingly, an alternative feature is provided to develop the NIRS-BCI system.

Stability analysis for Internet based teleoperated robot using prediction control

The variable time delay and the packet loss degrade the performance of Internet based teleoperation system seriously, even make the system unstable. Building upon the results of our recent work in [1], a provably stable event-prediction based control strategy is proposed for variable delay teleoperation. A Path Governor(PG) at master site, Generalized Predictive Controller (GPC) at slave site and a model of variable delay within a predictive control frame work are used to improve the response transparence. A Sparse Multivariable Linear Regression (SMLR) algorithm is proposed to predict the next Round Trip Timedelay(RTT). According to the next RTT, the PG is designed to generate the future event which can be used to compute the robotic position that is the one when the command was transmitted to the slave robot. The GPC can generate the redundant control information to diminish the influence of the packet loss and the large time delay in the internet to the system. Finally, a Lypunov-based analysis of the performance and stability of the resulting system is presented. Experiment results with a wheeled robot teleoperation setup demonstrate that these strategies can dynamically compensate for the variable time delay and reduce the performance degradation induced by packet loss.

Adaptive backstepping control of flexible joint robots with friction compensation based on LuGre model

Flexibility, parameter uncertainty and joint friction restrict the control performance of flexible joint robots. In this paper, an adaptive backstepping method with friction compensation is presented for the control of flexible joint robots with parameter inaccuracies, in which the joint friction is described by LuGre model and compensated base on a friction observer, and the inaccuracies of robot parameters are resolved by adaptive parameters estimation. Simulation verifies the effectiveness of the friction compensation method, and shows the proposed controller could achieve good control performance with model inaccuracies.

Improving Classification by Feature Discretization and Optimization for fNIRS-based BCI

In this paper, we present a signal discretization and feature selection method to improve classification accuracy for fNIRS based brain computer interface (BCI) system, which can classifiy right hand clench force motor imagery and clench speed motor imagery at an accuracy of 69%-81% through 5 fold cross validation in 6 subjects. Difference between oxyhemoglobin and deoxyhemoglobin (we abbreviate this difference as HbD) is proposed as a new feature type and shows outstanding performance in some subjects. Linear kernal support vector machine (SVM) classification between clench force motor imagery and clench speed motor imagery using four concentration feature types (oxyhemoglobin, deoxyhemoglobin, totalhemoglobin, and HbD) is implemented. Our results demonstrate that feature discretization using Chi2 algorighm and feature optimization using ‘MIFS’ (Mutual Information Feature Selection) criterion can improve the classification accuracy by more than 35%. Except total hemoglobin, all the other three feature types can be used as the optimum feature for different subjects. The results of this paper can also be used in online BCI applications.

Signal reconstruction of the slow wave and spike potential from electrogastrogram.

The gastric slow wave and the spike potential can correspondingly represent the rhythm and the intensity of stomach motility. Because of the filtering effect of biological tissue, electrogastrogram (EGG) cannot measure the spike potential on the abdominal surface in the time domain. Thus, currently the parameters of EGG adopted by clinical applications are only the characteristics of the slow wave, such as the dominant frequency, the dominant power and the instability coefficients. The limitation of excluding the spike potential analyses hinders EGG from being a diagnosis to comprehensively reveal the motility status of the stomach. To overcome this defect, this paper a) presents an EGG reconstruction method utilizing the specified signal components decomposed by the discrete wavelet packet transform, and b) obtains a frequency band for the human gastric spike potential through fasting and postprandial cutaneous EGG experiments for twenty-five human volunteers. The results indicate the lower bound of the human gastric spike potential frequency is 0.96±0.20 Hz (58±12 cpm), and the upper bound is 1.17±0.23 Hz (70±14 cpm), both of which have not been reported before to the best of our knowledge. As an auxiliary validation of the proposed method, synchronous serosa-surface EGG acquisitions are carried out for two dogs. The frequency band results for the gastric spike potential of the two dogs are respectively 0.83-0.90 Hz (50-54 cpm) and 1.05-1.32 Hz (63-79 cpm). They lie in the reference range 50-80 cpm proposed in previous literature, showing the feasibility of the reconstruction method in this paper.

Enhanced Performance by Time-Frequency-Phase Feature for EEG-Based BCI Systems

We introduce a new motor parameter imagery paradigm using clench speed and clench force motor imagery. The time-frequency-phase features are extracted from mu rhythm and beta rhythms, and the features are optimized using three process methods: no-scaled feature using “MIFS” feature selection criterion, scaled feature using “MIFS” feature selection criterion, and scaled feature using “mRMR” feature selection criterion. Support vector machines (SVMs) and extreme learning machines (ELMs) are compared for classification between clench speed and clench force motor imagery using the optimized feature. Our results show that no significant difference in the classification rate between SVMs and ELMs is found. The scaled feature combinations can get higher classification accuracy than the no-scaled feature combinations at significant level of 0.01, and the “mRMR” feature selection criterion can get higher classification rate than the “MIFS” feature selection criterion at significant level of 0.01. The time-frequency-phase feature can improve the classification rate by about 20% more than the time-frequency feature, and the best classification rate between clench speed motor imagery and clench force motor imagery is 92%. In conclusion, the motor parameter imagery paradigm has the potential to increase the direct control commands for BCI control and the time-frequency-phase feature has the ability to improve BCI classification accuracy.

Processing and analysis of bio-signals from human stomach

In this article, the electrogastrography is utilized to detect slow wave of gastric digest motility after test meal. In order to extract useful information, this study used multi-resolution method with the Daubechies wavelet function to decompose EGG signal into 9 layers. We reconstructed the slow wave with decomposed signal after digital signal processing to achieve method of the slow wave detection of EGG. During strong contraction of stomach, there is a significant increase in frequency spectrum and power spectrum of the slow wave frequency region. And power spectrum of time windows of slow wave bandwidth increases clearly. The contribution of this paper was that the filter of CWT and Fourier transform was used to obtain the bandwidth of slow wave, and the proposed method was compared with Chebyshev filter. By calculation and analysis of experimental data, the EGG slow wave detection method of wavelet-based motility of gastric digestion was verified to be effective, and also provided a better clinical method to monitor the state of stomach activities. This method is also can be applied to human medical sensor network which includes electrogastrography, electrocardiogram, thermometer, sphygmomanometer.

Imagined Hand Clenching Force and Speed Modulate Brain Activity and Are Classified by NIRS Combined With EEG

Simultaneous acquisition of brain activity signals from the sensorimotor area using NIRS combined with EEG, imagined hand clenching force and speed modulation of brain activity, as well as 6-class classification of these imagined motor parameters by NIRS-EEG were explored. Near infrared probes were aligned with C3 and C4, and EEG electrodes were placed midway between the NIRS probes. NIRS and EEG signals were acquired from six healthy subjects during six imagined hand clenching force and speed tasks involving the right hand. The results showed that NIRS combined with EEG is effective for simultaneously measuring brain activity of the sensorimotor area. The study also showed that in the duration of (0, 10) s for imagined force and speed of hand clenching, HbO first exhibited a negative variation trend, which was followed by a negative peak. After the negative peak, it exhibited a positive variation trend with a positive peak about 6–8 s after termination of imagined movement. During (−2, 1) s, the EEG may have indicated neural processing during the preparation, execution, and monitoring of a given imagined force and speed of hand clenching. The instantaneous phase, frequency, and amplitude feature of the EEG were calculated by Hilbert transform; HbO and the difference between HbO and Hb concentrations were extracted. The features of NIRS and EEG were combined to classify three levels of imagined force [at 20/50/80% MVGF (maximum voluntary grip force)] and speed (at 0.5/1/2 Hz) of hand clenching by SVM. The average classification accuracy of the NIRS-EEG fusion feature was 0.74 ± 0.02. These results may provide increased control commands of force and speed for a brain-controlled robot based on NIRS-EEG.