Christopher A. Russell

Removal of ocular artifacts from electro-encephalogram by adaptive filtering

The electro-encephalogram (EEG) is useful for clinical diagnosts and in biomedical research. EEG signals, however, especially those recorded from frontal channels, often contain strong electro-oculogram (EOG) artifacts produced by eye movements. Existing regression-based methods for removing EOG artifacts require various procedures for preprocessing and calibration that are inconvenient and timeconsuming. The paper describes a method for removing ocular artifacts based on adaptive filtering. The method uses separately recorded vertical EOG and horizontal EOG signals as two reference inputs. Each reference input is first processed by a finite impulse response filter of length M (M=3 in this application) and then subtracted from the original EEG. The method is implemented by a recursive leastsquares algorithm that includes a forgetting factor (λ=0.9999 in this application) to track the non-stationary portion of the EOG signals. Results from experimental data demonstrate that the method is easy to implement and stable, converges fast and is suitable for on-line removal of EOG artifacts. The first three coefficients (up to M=3) were significantly larger than any remaining coefficients.

Performance Enhancement in an Uninhabited Air Vehicle Task Using Psychophysiologically Determined Adaptive Aiding

Objective: We show that psychophysiologically driven real-time adaptive aiding significantly enhances performance in a complex aviation task. Afurther goal was to assess the importance of individual operator capabilities when providing adaptive aiding. Background: Psychophysiological measures are useful for monitoring cognitive workload in laboratory and real-world settings. They can be recorded without intruding into task performance and can be analyzed in real time, making them candidates for providing operator functional state estimates. These estimates could be used to determine if and when system intervention should be provided to assist the operator to improve system performance. Methods: Adaptive automation was implemented while operators performed an uninhabited aerial vehicle task. Psychophysiological data were collected and an artificial neural network was used to detect periods of high and low mental workload in real time. The high-difficulty task levels used to initiate the adaptive automation were determined separately for each operator, and a group-derived mean difficulty level was also used. Results: Psychophysiologically determined aiding significantly improved performance when compared with the no-aiding conditions. Improvement was greater when adaptive aiding was provided based on individualized criteria rather than on group-derived criteria. The improvements were significantly greater than when the aiding was randomly provided. Conclusion: These results show that psychophysiologically determined operator functional state assessment in real time led to performance improvement when included in closed loop adaptive automation with a complex task. Application: Potential future applications of this research include enhanced workstations using adaptive aiding that would be driven by operator functional state.

Operator Functional State Classification Using Multiple Psychophysiological Features in an Air Traffic Control Task

We studied 2 classifiers to determine their ability to discriminate among 4 levels of mental workload during a simulated air traffic control task using psychophysiological measures. Data from 7 air traffic controllers were used to train and test artificial neural network and stepwise discriminant classifiers. Very high levels of classification accuracy were achieved by both classifiers. When the 2 task difficulty manipulations were tested separately, the percentage correct classifications were between 84% and 88%. Feature reduction using saliency analysis for the artificial neural networks resulted in a mean of 90% correct classification accuracy. Considering the data as a 2-class problem, acceptable load versus overload, resulted in almost perfect classification accuracies, with mean percentage correct of 98%. In applied situations, the most important distinction among operator functional states would be to detect mental overload situations. These results suggest that psychophysiological data are capable of such discriminations with high levels of accuracy. Potential applications of this research include test and evaluation of new and modified systems and adaptive aiding.

The effects of day-to-day variability of physiological data on operator functional state classification.

The application of pattern classification techniques to physiological data has undergone rapid expansion. Tasks as varied as the diagnosis of disease from magnetic resonance images, brain-computer interfaces for the disabled, and the decoding of brain functioning based on electrical activity have been accomplished quite successfully with pattern classification. These classifiers have been further applied in complex cognitive tasks to improve performance, in one example as an input to adaptive automation. In order to produce generalizable results and facilitate the development of practical systems, these techniques should be stable across repeated sessions. This paper describes the application of three popular pattern classification techniques to EEG data obtained from asymptotically trained subjects performing a complex multitask across five days in one month. All three classifiers performed well above chance levels. The performance of all three was significantly negatively impacted by classifying across days; however two modifications are presented that substantially reduce misclassifications. The results demonstrate that with proper methods, pattern classification is stable enough across days and weeks to be a valid, useful approach.

Selection of input features across subjects for classifying crewmember workload using artificial neural networks

The issue of crewmember workload is important in complex system operation because operator overload leads to decreased mission effectiveness. Psychophysiological research on mental workload uses measures such as electroencephalogram (EEG), cardiac, eye-blink, and respiration measures to identify mental workload levels. This paper reports a research effort whose primary objective was to determine if one parsimonious set of salient psychophysiological features can be identified to accurately classify mental workload levels across multiple test subjects performing a multiple task battery. To accomplish this objective, a stepwise multivariate discriminant analysis heuristic and artificial neural network feature selection with a signal-to-noise ratio (SNR) are used. In general, EEG power in the 31-40-Hz frequency range and ocular input features appeared highly salient. The second objective was to assess the feasibility of a single model to classify mental workload across different subjects. A classification accuracy of 87% was obtained for seven independent validation subjects using neural network models trained with data from other subjects. This result provides initial evidence for the potential use of generalized classification models in multitask workload assessment.

Performance Enhancement With Real-Time Physiologically Controlled Adaptive Aiding

The realization of optimal system performance is the goal of both system designers and users. One critical component in attaining this goal is proper operator functioning. In contemporary systems the functional state of the operator is not considered during system operation. Degraded states of operator functioning can result from the demands of controlling complex systems, the work environment and internal operator variables. This, in turn, can lead to errors and overall suboptimal system performance. In the case of mental workload, system performance could be improved by reducing task demands during periods of operator overload. Accurate estimation of the operators functional state is crucial to successful implementation of an adaptive aiding system. One method of determining operator functional state is by monitoring the operators physiology. In the present study, physiological signals were used to continuously monitor subjects functional state and to adapt the task by reducing the number of subtasks when high levels of mental workload were detected. The goal was to demonstrate performance improvement with adaptive aiding. Because adaptive aiding during high mental workload has not been previously implemented its benefit has not be demonstrated. Application of adaptive aiding techniques reduced tracking task error by 44% and resource monitoring error by 33%. These results demonstrate the utility of adaptive aiding using physiological measures with artificial neural networks to determine the appropriate time to introduce the aiding.

Removal of Ocular Artifacts from EEG: A Comparison of Adaptive Filtering Method and Regression Method Using Simulated Data

We recently proposed an adaptive filtering method for removing ocular artifacts from EEG recordings. In this study, the accuracy of this method is evaluated quantitatively using simulated data and compared with the accuracy of the time domain regression method. The results show that when transfer of ocular signal to EEG channel is frequency dependent, or when there is a time delay, the adaptive filtering method is more accurate in recovering the true EEG signals

Operator Functional State Classification Using Neural Networks with Combined Physiological and Performance Features

Operator functional state assessment is a critical component of adaptive aiding systems. A combination of physiological and performance variables were used with a neural network to determine operator functional state. A multiple task battery provided three levels of mental workload. The data were randomly divided into two data sets, one was used to train a neural net and the other to test the accuracy of the trained neural net. The results showed an overall correct classification of 86.8% for the test data set. For the three levels of task difficulty the correct classification was 90.5% for low, 81.7% for the medium and 88.3% for the high. These results support the use of combined physiological and performance data to obtain high levels of operator functional state classification accuracy. The optimal approach to utilizing this information during system operation will have to be developed. With current technology the development of small, wearable operator state classifiers is possible.

The Importance of Determining Individual Operator Capabilities When Applying Adaptive Aiding

Performance was significantly improved in an Uninhabited Air Vehicle (UAV) task when individually determined task difficulty levels were used to present psychophysiologically controlled adaptive aiding. Previous research in our laboratory demonstrated that the benefit of adaptive aiding varied according to the operators skill level when a common task difficulty was used for all operators. In the present study the difficult task level was determined for each operator. An average task difficulty level was also used. The best performance occurred when adaptive aiding was presented based upon psychophysiological data submitted to an artificial neural network when the implementation level was individually determined for each operator. Performance improvement using the mean difficulty level was lower as were the results when the adaptive aiding was randomly presented. Individual cognitive capability must be considered to achieve optimal performance via adaptive aiding.

Psychophysiologically Determined Classification of Cognitive Activity

Psychophysiological measures and artificial neural networks were used to determine how well higher levels of cognitive activity, such as executive function, spatial and verbal working memory and global workload, could be assessed. A complex uninhabited air vehicle simulator was used in which subjects were responsible for four vehicles simultaneously. The subjects had to evaluate visual images and maintain the status of the vehicles. The results showed that the cognitive states, derived from subjective reports, could be accurately classified. These results have application in human factors environments which demand higher level cognitive processing and may be useful when implementing adaptive aiding in these systems.