Abstract: We propose an evolutionary state-space model (E-SSM) for analyzing high-dimensional brain signals, the statistical properties of which evolve over the course of a nonspatial memory experiment. Under the E-SSM, brain signals are modeled as mixtures of components (e.g., an AR(2) process) with oscillatory activity at predefined frequency bands. To account for the potential nonstationarity of these components (because brain responses can vary throughout an experiment), the parameters are allowed to vary over epochs. Compared with classical approaches, such as independent component analyses and filtering, the proposed method accounts for the entire temporal correlation of the components and accommodates nonstationarity. For inference purposes, we propose a novel computational algorithm based on a Kalman smoother, maximum likelihood, and blocked resam- pling. The E-SSM model is applied in simulation studies and applied to multi-epoch local field potential (LFP) signal data, collected from a nonspatial (olfactory) sequence memory task study. The results confirm that our method captures the evolution of the power of the components across different phases in the experiment, and identifies clusters of electrodes that behave similarly with respect to the decomposition of different sources. These findings suggest that the activity of electrodes does change over the course of an experiment in practice. Thus, treating these epoch recordings as realizations of an identical process could lead to misleading results. In summary, the proposed method underscores the importance of capturing the evolution in brain responses over the study period.

Key words and phrases: Auto-regressive model, brain signals, state-space models, spectral analysis, time-frequency analysis.