WP2. Sensor configuration & signal capturing

Description of work

T2.1 – Configuration, calibration and noise-reduction in eye-tracking movements (SMI, UNI KO-LD) [M1-M7, M24-M32]

The goal of this task is to specify the parameters of the eye-tracker that best fits to MAMEM’s requirements. Though the eye tracking techniques available today are highly sophisticated, they are still far from perfect. We assume that it will be even more difficult to track the eyes of people who have lost their fine motor skills, because this may also affect them positioning their head or they may even shake involuntarily. Therefore, in this task we will adopt a novel approach for eye-tracking where the signals recorder by the eye tracking glasses, i.e. eye fixations and saccades, are accurately mapped to the corresponding point from the display, which is looked upon by the user. The advantage of glasses is that it is easier to track the eyes, but it requires additional synchronization effort then to find which spot on the display is monitored. The sampling rate will be chosen to be minimal regarding the computation cost (real-time functionality), but yet the highest possible to ensure high quality information. Over-sensitive trackers lose calibration far too soon. The compromise between efficient eye-tracking and the time spent for calibration will serve as the basis for the configuration of the eye-tracker.

T2.2 – Configuration, calibration and noise-reduction in EEG signals (EBNeuro, CERTH) [M1-M7, M24-M32]

The goal of this task is to define the EEG recorder specifications that will facilitate the process of translating the captured brain electrical signals into mental commands. In the first stage the general configuration will be addressed, while in the second stage the initial configuration will be reviewed incorporating the results of the first phase pilot trials performed in WP6. Overall, the EEG recorder has to be as less cumbersome as possible and the preparation time has to be decreased towards an easy-to-use BCI system. The exact type of electrodes (i.e., wet vs. dry, saline vs. gel) will be explored towards efficiently capturing EEG-signals with unobtrusive and non-invasive solutions. Other parameters, i.e., the sampling rate and the electrodes number and positioning will be investigated based on the requirements of the problem at hand. The particularities of user disabilities will play a vital role in deciding on these parameters and the overall configuration. For instance, the EEG signal contamination induced by head tremor (i.e., a common symptom in PD) will have to be reduced, e.g., by applying bandpass and notch filtering to the recorder signals. The parameters and the type of filters in this pre-processing phase will be selected based on the trade-off between high performance and real-time functionality. Similarly, the noise induced by cable movement will have to be eliminated, e.g., by using a wireless configuration provided that the bandwidth is sufficient. Moreover, the use of active electrodes will be explored as they help to accomplish high signal-to-noise ratio, which is extremely important for ‘out-of-lab’ functionality. The module configuration will be further tailored to account for user’s comfort and usability issues. Finally, the time required for calibration will be also considered in deciding the optimal configuration setting.

T2.3 – Configuration, calibration and noise-reduction in bio-measurements (SMI, CERTH) [M1-M7, M24-M32]

The goal of this task is first to select the most appropriate bio-measurements and then to specify the configuration and calibration procedure that is necessary for tuning the received signal according to the requirements of our system. Sensors providing galvanic skin response (GSR) and heart rate (HR) measurements will be evaluated and the most appropriate will be selected in terms of balancing the trade-off between the effectiveness in capturing the user’s emotional context (i.e., stress level) and minimum calibration time.  The bio-measurement module will be selected to be adaptive to user particularities, i.e., the module will be able to operate based on the GSR or the HR alone or by combining these measurements.

T2.4 – Optimizations with respect to near real-time, large scale and synchronized access to signals (EBNeuro, SMI, CERTH, UNI KO-LD) [M8-M18, M27-M32]

The goal of this task is to bring together all independent sensor modules designed in the previous tasks (T2.1-T2.3). The overall sensor installation will be implemented herein taking into account real-time and large-scale data manipulation. The parameters of each module will be optimized in an attempt to facilitate synchronous recording and data processing. In this context, the requirements imposed by the technical aspect of data acquisition (i.e., limitations in sampling rate of eye-tracking and EEG recording) will be jointly considered, taking also into account the requirement to turn the sampled signals into smooth interface operations. For instance, in order to translate a specific wave pattern of eye movements into a movement of the display pointer we might need a sampling rate for the eye tracker much higher than the typical mouse sampling rate.

With respect to synchronization, given the anticipated differences in the sampling rates of the different sensors (i.e., eye tracker, EEG recorder, bio-sensors) the use of the appropriate strategy for time-stamping the received signal will be essential for their consistent alignment. In this respect, we plan to consider the adoption of an asynchronous strategy for acquiring the signals where each signal is being sampled at its own rate. Then, precise time stamps are associated to every data chunk which can be used for the time-alignment of the signals and their synchronized processing. Equally important are also the strategies adopted for reading (e.g., blocking/unblocking reads, re-sampling), storing and representing the data so as to be able to facilitate the data flow that fits the needs of the application at hand (e.g., recording or real time processing). The strategies that we plan to investigate have been successfully implemented by RTMaps[1], which is a component based framework for real-time, multi-modal applications. Finally, T2.4 will also investigate the design of an auto-calibration setup for the overall sensor installation, requiring minimal effort and no expert knowledge. If necessary, this auto-calibration setup will be complemented by the design and implementation of a series of calibration and fine-tuning processes for the adaptation of the system to the requirements of a home environment. The final goal is the development of a lightweight installation for the necessary sensors that will be easily portable and appropriate for non-controlled environments.

[1] http://www.intempora.com/