Improving user experience of SSVEP BCI through low amplitude depth and high frequency stimuli design.

Steady-States Visually Evoked Potentials (SSVEP) refer to the sustained rhythmic activity observed in surface electroencephalography (EEG) in response to the presentation of repetitive visual stimuli (RVS). Due to their robustness and rapid onset, SSVEP have been widely used in Brain Computer Interfaces (BCI). However, typical SSVEP stimuli are straining to the eyes and present risks of triggering epileptic seizures. Reducing visual stimuli contrast or extending their frequency range both appear as relevant solutions to address these issues. It however remains sparsely documented how BCI performance is impacted by these features and to which extent user experience can be improved. We conducted two studies to systematically characterize the effects of frequency and amplitude depth reduction on SSVEP response. The results revealed that although high frequency stimuli improve visual comfort, their classification performance were not competitive enough to design a reliable/responsive BCI. Importantly, we found that the amplitude depth reduction of low frequency RVS is an effective solution to improve user experience while maintaining high classification performance. These findings were further validated by an online T9 SSVEP-BCI in which stimuli with 40% amplitude depth reduction achieved comparable results (>90% accuracy) to full amplitude stimuli while significantly improving user experience.

Inattentional deafness to auditory alarms: Inter-individual differences, electrophysiological signature and single trial classification.

Inattentional deafness can have deleterious consequences in complex real-life situations (e.g. healthcare, aviation) leading to miss critical auditory signals. Such failure of auditory attention is thought to rely on top-down biasing mechanisms at the central executive level. A complementary approach to account for this phenomenon is to consider the existence of visual dominance over hearing that could be implemented via direct visual-to-auditory pathways. To investigate this phenomenon, thirteen aircraft pilots, equipped with a 32-channel EEG system, faced a low and high workload scenarii along with an auditory oddball task in a motion flight simulator. Prior to the flying task, the pilots were screened to assess their working memory span and visual dominance susceptibility. The behavioral results disclosed that the volunteers missed 57.7% of the auditory alarms in the difficult condition. Among all evaluated capabilities, only the visual dominance index was predictive of the miss rate in the difficult scenario. These findings provide behavioral evidences that other early cross-modal competitive process than top down modulation process could account for inattentional deafness. The electrophysiological analyses showed that the miss over the hit alarms led to a significant amplitude reduction of early perceptual (N100) and late attentional (P3a and P3b) event-related potentials components. Eventually, we implemented an EEG-based processing pipeline to perform single-trial classification of inattentional deafness. The results indicate that this processing chain could be used in an ecological setting as it led to 72.2% mean accuracy to discriminate missed from hit auditory alarms.

Using near infrared spectroscopy and heart rate variability to detect mental overload.

Mental workload is a key factor influencing the occurrence of human error, especially during piloting and remotely operated vehicle (ROV) operations, where safety depends on the ability of pilots to act appropriately. In particular, excessively high or low mental workload can lead operators to neglect critical information. The objective of the present study is to investigate the potential of functional near infrared spectroscopy (fNIRS) - a non-invasive method of measuring prefrontal cortex activity - in combination with measurements of heart rate variability (HRV), to predict mental workload during a simulated piloting task, with particular regard to task engagement and disengagement. Twelve volunteers performed a computer-based piloting task in which they were asked to follow a dynamic target with their aircraft, a task designed to replicate key cognitive demands associated with real life ROV operating tasks. In order to cover a wide range of mental workload levels, task difficulty was manipulated in terms of processing load and difficulty of control - two critical sources of workload associated with piloting and remotely operating a vehicle. Results show that both fNIRS and HRV are sensitive to different levels of mental workload; notably, lower prefrontal activation as well as a lower LF/HF ratio at the highest level of difficulty, suggest that these measures are suitable for mental overload detection. Moreover, these latter measurements point toward the existence of a quadratic model of mental workload.