In this paper, we present a novel approach to solving the single-trial evoked-potential estimation problem. Recognizing that different components of an evoked potential complex may originate from different functional brain sites and can be distinguished according to their respective latencies and amplitudes, we propose an estimation approach based on identification of evoked potential components on a single-trial basis. The estimation process is performed in two stages: first, an average evoked potential is calculated and decomposed into a set of components, with each component serving as a subtemplate for the next stage; then, the single measurement is parametrically modeled by a superposition of an emulated ongoing electroencephalographic activity and a linear combination of latency and amplitude-corrected component templates. Once optimized, the model provides the two assumed signal contributions, namely the ongoing brain activity and the single evoked brain response. The estimator's performance is analyzed analytically and via simulation, verifying its capability to extract single components at low signal-to-noise ratios typical of evoked potential data. Finally, two applications are presented, demonstrating the improved analysis capabilities gained by using the proposed approach. The first application deals with movement related brain potentials, where a change of the single evoked response due to external loading is detected. The second application involves cognitive event-related brain potentials, where a dynamic change of two overlapping components throughout the experimental session is detected and tracked.

Current estimators for single-trial evoked potentials (EP's) require a signal-to-noise ratio (SNR) of 0 dB or better to obtain high quality estimations, yet many types of EP's suffer from substantially lower SNR's. This paper presents a robust-evoked-potential-estimator (REPE) facilitating high quality estimations of single movement related EP's with a relatively low SNR. The estimator is based on a standard ARX model, enhanced to support estimation under poor SNR conditions. The REPE was tested successfully on a computer simulated data set giving reliable single-trial estimations for the low SNR range of around -20 dB. THe REPE was also applied to experimental data, producing clear single-trial estimations of movement related brain signals recorded in a classic scenario of self-paced finger tapping experiment.

No abstract

No abstract

We present a novel approach to the problem of event-related potential (ERP) identification, based on a competitive artificial neural network (ANN) structure. Our method uses ensembled electroencephalogram (EEG) data just as used in conventional averaging, however without the need for a priori data subgrouping into distinct categories (e.g., stimulus- or event-related), and thus avoids conventional assumptions on response invariability. The competitive ANN, often described as a winner takes all neural structure, is based on dynamic competition among the net neurons where learning takes place only with the winning neuron. Using a simple single-layered structure, the proposed scheme results in convergence of the actual neural weights to the embedded ERP patterns. The method is applied to real event-related potential data recorded during a common odd-ball type paradigm. For the first time, within-session variable signal patterns are automatically identified, dismissing the strong and limiting requirement of a priori stimulus-related selective grouping of the recorded data. The results present new possibilities in ERP research.