Action prediction based on anticipatory brain potentials during simulated driving

Objective. The ability of an automobile to infer the driver's upcoming actions directly from neural signals could enrich the interaction of the car with its driver. Intelligent vehicles fitted with an on-board brain–computer interface able to decode the driver's intentions can use this information to improve the driving experience. In this study we investigate the neural signatures of anticipation of specific actions, namely braking and accelerating. Approach. We investigated anticipatory slow cortical potentials in electroencephalogram recorded from 18 healthy participants in a driving simulator using a variant of the contingent negative variation (CNV) paradigm with Go and No-go conditions: count-down numbers followed by 'Start'/'Stop' cue. We report decoding performance before the action onset using a quadratic discriminant analysis classifier based on temporal features. Main results. (i) Despite the visual and driving related cognitive distractions, we show the presence of anticipatory event related potentials locked to the stimuli onset similar to the widely reported CNV signal (with an average peak value of −8 μV at electrode Cz). (ii) We demonstrate the discrimination between cases requiring to perform an action upon imperative subsequent stimulus (Go condition, e.g. a 'Red' traffic light) versus events that do not require such action (No-go condition; e.g. a 'Yellow' light); with an average single trial classification performance of 0.83 ± 0.13 for braking and 0.79 ± 0.12 for accelerating (area under the curve). (iii) We show that the centro-medial anticipatory potentials are observed as early as 320 ± 200 ms before the action with a detection rate of 0.77 ± 0.12 in offline analysis. Significance. We show for the first time the feasibility of predicting the driver's intention through decoding anticipatory related potentials during simulated car driving with high recognition rates.

EEG-based decoding of error-related brain activity in a real-world driving task

Ricardo Andres Chavarriaga Lozano, Lucian Andrei Gheorghe, Inaki Asier Iturrate Gil, Zahra Khaliliardali, Huaijian Zhang

Objectives. Recent studies have started to explore the implementation of brain–computer interfaces (BCI) as part of driving assistant systems. The current study presents an EEG-based BCI that decodes error-related brain activity. Such information can be used, e.g., to predict driver's intended turning direction before reaching road intersections. Approach. We executed experiments in a car simulator (N = 22) and a real car (N = 8). While subject was driving, a directional cue was shown before reaching an intersection, and we classified the presence or not of an error-related potentials from EEG to infer whether the cued direction coincided with the subject's intention. In this protocol, the directional cue can correspond to an estimation of the driving direction provided by a driving assistance system. We analyzed ERPs elicited during normal driving and evaluated the classification performance in both offline and online tests. Results. An average classification accuracy of 0.698 ± 0.065 was obtained in offline experiments in the car simulator, while tests in the real car yielded a performance of 0.682 ± 0.059. The results were significantly higher than chance level for all cases. Online experiments led to equivalent performances in both simulated and real car driving experiments. These results support the feasibility of decoding these signals to help estimating whether the driver's intention coincides with the advice provided by the driving assistant in a real car. Significance. The study demonstrates a BCI system in real-world driving, extending the work from previous simulated studies. As far as we know, this is the first online study in real car decoding driver's error-related brain activity. Given the encouraging results, the paradigm could be further improved by using more sophisticated machine learning approaches and possibly be combined with applications in intelligent vehicles.

Institute of Physics2015

Combining Discriminant and Topographic Information in BCI: Preliminary Results on Stroke Patients

Andrea Biasiucci, Ricardo Andres Chavarriaga Lozano, Benjamin Hamner, Robert Leeb

Non–Invasive Brain–Computer Interfaces (BCI) convey a great potential in the field of stroke rehabilitation, where the continuous monitoring of mental tasks execution could support the positive effects of standard therapies. In this paper we combine time-frequency analysis of EEG with the topographic analysis to identify and track task–related patterns of brain activity emerging during a single BCI session. 6 Stroke patients executed Motor Imagery of the affected and unaffected hands: EEG sites were ranked depending on their discriminant power (DP) at different time instants and the resulting discriminant periods were used as a prior to extract EEG Microstates. Results show that the combination of these two techniques can provide insights about specific motor–related processes happening at a fine grain temporal resolution. Such events, represented by EEG microstates, can be tracked and used both to quantify changes of underlying neural structures and to provide feedback to patients and therapists.

Ieee Service Center, 445 Hoes Lane, Po Box 1331, Piscataway, Nj 08855-1331 Usa2011

Bayesian machine learning applied in a brain-computer interface for disabled users

Ulrich Hoffmann

A brain-computer interface (BCI) is a system that enables control of devices or communication with other persons, only through cerebral activity, without using muscles. The main application for BCIs is assistive technology for disabled persons. Examples for devices that can be controlled by BCIs are artificial limbs, spelling devices, or environment control systems. BCI research has seen renewed interest in recent years, and it has been convincingly shown that communication via a BCI is in principle feasible. However, present day systems still have shortcomings that prevent their widespread application. In part, these shortcomings are caused by limitations in the functionality of the pattern recognition algorithms used for discriminating brain signals in BCIs. Moreover, BCIs are often tested exclusively with able-bodied persons instead of conducting tests with the target user group, namely disabled persons. The goal of this thesis is to extend the functionality of pattern recognition algorithms for BCI systems and to move towards systems that are helpful for disabled users. We discuss extensions of linear discriminant analysis (LDA), which is a simple but efficient method for pattern recognition. In particular, a framework from Bayesian machine learning, the so-called evidence framework, is applied to LDA. An algorithm is obtained that learns classifiers quickly, robustly, and fully automatically. An extension of this algorithm allows to automatically reduce the number of sensors needed for acquisition of brain signals. More specifically, the algorithm allows to perform electrode selection. The algorithm for electrode selection is based on a concept known as automatic relevance determination (ARD) in Bayesian machine learning. The last part of the algorithmic development in this thesis concerns methods for computing accurate estimates of class probabilities in LDA-like classifiers. These probabilities are used to build a BCI that dynamically adapts the amount of acquired data, so that a preset, approximate bound on the probability of misclassifications is not exceeded. To test the algorithms described in this thesis, a BCI specifically tailored for disabled persons is introduced. The system uses electroencephalogram (EEG) signals and is based on the P300 evoked potential. Datasets recorded from five disabled and four able-bodied subjects are used to show that the Bayesian version of LDA outperforms plain LDA in terms of classification accuracy. Also, the impact of different static electrode configurations on classification accuracy is tested. In addition, experiments with the same datasets demonstrate that the algorithm for electrode selection is computationally efficient, yields physiologically plausible results, and improves classification accuracy over static electrode configurations. The classification accuracy is further improved by dynamically adapting the amount of acquired data. Besides the datasets recorded from disabled and able-bodied subjects, benchmark datasets from BCI competitions are used to show that the algorithms discussed in this thesis are competitive with state-of-the-art electroencephalogram (EEG) classification algorithms. While the experiments in this thesis are uniquely performed with P300 datasets, the presented algorithms might also be useful for other types of BCI systems based on the EEG. This is the case because functionalities such as robust and automatic computation of classifiers, electrode selection, and estimation of class probabilities are useful in many BCI systems. Seen from a more general point of view, many applications that rely on the classification of cerebral activity could possibly benefit from the methods developed in this thesis. Among the potential applications are interrogative polygraphy ("lie detection") and clinical applications, for example coma outcome prognosis and depth of anesthesia monitoring.

EPFL2007