Javier Herrera-Vega

Posture Based Detection of Attention in Human Computer Interaction

Unacted posture conveys cues about people’s attentional disposition. We aim to identify robust markers of attention from posture while people carry out their duties seated in front of their computers at work. Body postures were randomly captured from 6 subjects while at work using a Kinect, and self-assessed as attentive or not attentive. Robust postural features exhibiting higher discriminative power across classification exercises with 4 well-known classifiers were identified. Average classification of attention from posture reached 76.47%±4.58% (F-measure). A total of 40 postural features were tested and those proxy of head tilt were found to be the most stable markers of attention in seated conditions based upon 3 class separability criteria. Unobtrusively monitoring posture of users while working in front of a computer can reliably be used to infer attentional disposition from the user. Human-computer interaction systems can benefit from this knowledge to customize the experience to the user changing attentional state.

MOCARTS: A lightweight radiation transport simulator for easy handling of complex sensing geometries

In functional neuroimaging (fNIRS), elaborated sensing geometries pairing multiple light sources and detectors arranged over the tissue surface are needed. A variety of software tools for probing forward models of radiation transport in tissue exist, but their handling of sensing geometries and specification of complex tissue architectures is, most times, cumbersome. In this work, we introduce a lightweight simulator, Monte Carlo Radiation Transport Simulator (MOCARTS) that attends these demands for simplifying specification of tissue architectures and complex sensing geometries. An object-oriented architecture facilitates such goal. The simulator core is evolved from the Monte Carlo Multi-Layer (mcml) tool but extended to support multi-channel simulations. Verification against mcml yields negligible error (RMSE~4-10e-9) over a photon trajectory. Full simulations show concurrent validity of the proposed tool. Finally, the ability of the new software to simulate multi-channel sensing geometries and to define biological tissue models in an intuitive nested-hierarchy way are exemplified.

Developing a device for monitoring O2 saturation in blood

We present a single-channel device for monitoring oxygen saturation in blood using near infrared light (NIR). This device measures the differential absorption of both oxi-hemoglobin (HbO2) and deoxi-hemoglobin (HHb) concentrations using wavelengths within the biophotonic window (700 nm – 1100 nm). Our device works with two wavelengths: the first source (λ1 = 632 nm) operates in continuous wave (CW) and the second (λ2 = 940 nm) operates in pulsed mode. The pulsed signal at λ2 operates in a 20% fill factor. The CW signal provides information related to the arterial pulse, whereas both CW and pulsed signal provide information about blood oxygenation. Light emitted from the sources travels through the tissue and are collected on the detector by transmission. Measurements of blood oxygenation measured in transmission on fingers tips are presented which are comparable to commercial devices. The values obtained are above 90.0% blood oxygenation in healthy tests subjects with 0.1% accuracy in the measurements.

A local multiscale probabilistic graphical model for data validation and reconstruction, and its application in industry

Abstract The detection and subsequent reconstruction of incongruent data in time series by means of observation of statistically related information is a recurrent issue in data validation. Unlike outliers, incongruent observations are not necessarily confined to the extremes of the data distribution. Instead, these rogue observations are unlikely values in the light of statistically related information. This paper proposes a multiresolution Bayesian network model for the detection of rogue values and posterior reconstruction of the erroneous sample for non-stationary time-series. Our method builds local Bayesian Network models that best fit to segments of data in order to achieve a finer discretization and hence improve data reconstruction. Our local multiscale approach is compared against its single-scale global predecessor (assumed as our gold standard) in the predictive power and of this, both error detection capabilities and error reconstruction capabilities are assessed. This parameterization and verification of the model are evaluated over three synthetic data source topologies. The virtues of the algorithm are then further tested in real data from the steel industry where the aforementioned problem characteristics are met but for which the ground truth is unknown. The proposed local multiscale approach was found to dealt better with increasing complexities in data topologies.

Modelling and validation of diffuse reflectance of the adult human head for fNIRS: scalp sub-layers definition.

Accurate estimation of brain haemodynamics parameters such as cerebral blood flow and volume as well as oxygen consumption i.e. metabolic rate of oxygen, with funcional near infrared spectroscopy (fNIRS) requires precise characterization of light propagation through head tissues. An anatomically realistic forward model of the human adult head with unprecedented detailed specification of the 5 scalp sublayers to account for blood irrigation in the connective tissue layer is introduced. The full model consists of 9 layers, accounts for optical properties ranging from 750nm to 950nm and has a voxel size of 0.5mm. The whole model is validated comparing the predicted remitted spectra, using Monte Carlo simulations of radiation propagation with 108 photons, against continuous wave (CW) broadband fNIRS experimental data. As the true oxy- and deoxy-hemoglobin concentrations during acquisition are unknown, a genetic algorithm searched for the vector of parameters that generates a modelled spectrum that optimally fits the experimental spectrum. Differences between experimental and model predicted spectra was quantified using the Root mean square error (RMSE). RMSE was 0.071 ± 0.004, 0.108 ± 0.018 and 0.235±0.015 at 1, 2 and 3cm interoptode distance respectively. The parameter vector of absolute concentrations of haemoglobin species in scalp and cortex retrieved with the genetic algorithm was within histologically plausible ranges. The new model capability to estimate the contribution of the scalp blood flow shall permit incorporating this information to the regularization of the inverse problem for a cleaner reconstruction of brain hemodynamics.