@INPROCEEDINGS{6944010,
author={Heger, D. and Herff, C. and Schultz, T.},
booktitle={Engineering in Medicine and Biology Society (EMBC), 2014 36th Annual International Conference of the IEEE},
title={Combining feature extraction and classification for fNIRS BCIs by regularized least squares optimization},
year={2014},
pages={2012-2015},
abstract={In this paper, we show that multiple operations of the typical pattern recognition chain of an fNIRS-based BCI, including feature extraction and classification, can be unified by solving a convex optimization problem. We formulate a regularized least squares problem that learns a single affine transformation of raw HbO2 and HbR signals. We show that this transformation can achieve competitive results in an fNIRS BCI classification task, as it significantly improves recognition of different levels of workload over previously published results on a publicly available n-back data set. Furthermore, we visualize the learned models and analyze their spatio-temporal characteristics.},
keywords={affine transforms;biochemistry;bioelectric potentials;brain-computer interfaces;electroencephalography;feature extraction;feature selection;infrared spectra;least mean squares methods;medical signal processing;molecular biophysics;optimisation;oxygen;proteins;spatiotemporal phenomena;O2;affine transformation;convex optimization problem;deoxygenated hemoglobin signals;fNIRS BCI classification task;feature classification;feature extraction;functional near-infrared spectroscopy;pattern recognition chain;regularized least squares optimization;spatiotemporal characteristic analysis;Analytical models;Brain models;Data models;Feature extraction;Optimization;Predictive models},
doi={10.1109/EMBC.2014.6944010},
ISSN={1557-170X},
month={Aug},}