We propose to enhance the decision making of pilot, co-pilot teams, over a range of vehicle platforms, with the aid
of neuroscience. The goal is to optimize this collaborative decision making interplay in time-critical, stressful
situations. We will research and measure human facial expressions, personality typing, and brainwave measurements to help answer questions related to optimum decision-making in group situations. Further, we propose to examine the nature of intuition in this decision making process. The brainwave measurements will be facilitated by a University of California, San Diego (UCSD) developed wireless Electroencephalography (EEG) sensing cap. We propose to measure brainwaves covering the whole head area with an electrode density of N=256, and yet keep
within the limiting wireless bandwidth capability of m=32 readouts. This is possible because solving Independent Component Analysis (ICA) and finding the hidden brainwave sources allow us to concentrate selective measurements with an organized sparse source →s sensing matrix [Φ<sub>s</sub>], rather than the traditional purely random compressive sensing (CS) matrix[Φ].
A new group analysis method to summarize the task-related BOLD responses based on independent
component analysis (ICA) was presented. As opposite to the previously proposed group ICA (gICA)
method, which first combined multi-subject fMRI data in either temporal or spatial domain and applied
ICA decomposition only once to the combined fMRI data to extract the task-related BOLD effects, the
method presented here applied ICA decomposition to the individual subjects' fMRI data to first find the
independent BOLD effects specifically for each individual subject. Then, the task-related independent
BOLD component was selected among the resulting independent components from the single-subject ICA
decomposition and hence grouped across subjects to derive the group inference. In this new ICA group
analysis (ICAga) method, one does not need to assume that the task-related BOLD time courses are
identical across brain areas and subjects as used in the grand ICA decomposition on the spatially
concatenated fMRI data. Neither does one need to assume that after spatial normalization, the voxels at
the same coordinates represent exactly the same functional or structural brain anatomies across different
subjects. These two assumptions have been problematic given the recent BOLD activation evidences.
Further, since the independent BOLD effects were obtained from each individual subject, the ICAga method can better account for the individual differences in the task-related BOLD effects. Unlike the gICA approach whereby the task-related BOLD effects could only be accounted for by a single unified BOLD model across multiple subjects. As a result, the newly proposed method, ICAga, was able to better fit the task-related BOLD effects at individual level and thus allow grouping more appropriate multisubject
BOLD effects in the group analysis.
This study explores the use of Independent Component Analysis (ICA) applied to normalized logarithmic spectral
changes in the activities of brain processes separated by spatial filters learned from electroencephalogram (EEG) data
using a temporal ICA. EEG data were collected during 1-2 hour virtual-reality based driving experiments, in which
subjects were instructed to maintain their cruising position and compensate for randomly induced drifts using the
steering wheel. ICA was first applied to 30-channel EEG data to separate the recorded signals into a sum of maximally
temporally independent components (ICs) for each of 15 subjects. Logarithmic spectra of IC activities were then
submitted to PCA-ICA to find spectrally fixed and temporally independent modulator (IM) processes. The second ICA
detected and modeled independent co-modulatory systems that multiplicatively affect the activities of spatially distinct
IC processes. Across subjects, we found two consistent temporally independent modulators: theta-beta and alpha
modulators that mediate spectral activations of the distinct cortical areas when the participants experience waves of
alternating alertness and drowsiness during long hour simulated driving. Furthermore, the time courses of the theta-beta
modulator were highly correlated with concurrent changes in subject driving error (a behavioral index of drowsiness).
Electroencephalograph (EEG) recording systems offer a versatile, noninvasive window on the brain's spatio-temporal activity for many neuroscience and clinical applications. Our research aims at improving the spatial
resolution and mobility of EEG recording by reducing the form factor, power drain and signal fanout of the
EEG acquisition node in a scalable sensor array architecture. We present such a node integrated onto a dimesized
circuit board that contains a sensor's complete signal processing front-end, including amplifier, filters,
and analog-to-digital conversion. A daisy-chain configuration between boards with bit-serial output reduces
the wiring needed. The circuit's low power consumption of 423 &mgr;W supports EEG systems with hundreds of
electrodes to operate from small batteries for many hours.
Coupling between the bit-serial output and the highly sensitive analog input due to dense integration of analog
and digital functions on the circuit board results in a deterministic noise component in the output, larger than
the intrinsic sensor and circuit noise. With software correction of this noise contribution, the system achieves
an input-referred noise of 0.277 &mgr;Vrms in the signal band of 1 to 100 Hz, comparable to the best medical-grade
systems in use. A chain of seven nodes using EEG dry electrodes created in micro-electrical-mechanical system
(MEMS) technology is demonstrated in a real-world setting.
SC715: Independent Component Analysis and Beyond: Blind Signal Processing and its Applications
Blind Signal Processing (BSP) is an emerging area of research and technology with solid theoretical foundations and many potential applications. The problems of separating or extracting of the source signals from sensor arrays, without knowledge of the transmission channel characteristics and the real sources, can be expressed briefly as a number of blind source separation (BSS) or related generalized component analysis (GCA) methods: Independent Component Analysis (ICA) (and its extensions), Sparse Component Analysis (SCA), Sparse Principal Component Analysis (SPCA), Non-negative Matrix Factorization (NMF), Time-Frequency Component Analyzer (TFCA) and Multichannel Blind Deconvolution (MBD). BSP is not limited to ICA or BSS. With BSP we aim to discover and validate principles or laws which govern relationships between inputs (hidden components) and outputs (observations) when the information about the propagation Multi-Input Multi-Output (MIMO) system and its inputs are limited or hindered. BSP incorporates many problems, like blind identification of channels of unknown systems or a problem of suitable decomposition of signals into basic latent (hidden) components which do not necessary represent true sources but rather some of their features or sub-components.
This four-hour course presents the fundamentals of blind signal processing, especially blind source separation and extraction, and in the remaining time discusses their applications in several important signal processing areas including estimation of sources, novel enhancement, denoising, artifact removal, filtering, detection, classification of multi-sensory signals and data, especially in biomedical applications and Brain Computer Interface (BCI).