Space-Time Correlations of EEG Time-Series
Haocheng Ni, Yue Yu, Pavel Loskot · 2022 2nd International Conference on Computer Science, Electronic Information Engineering and Intelligent Control Technology (CEI) · 2022
The spatially distributed stochastic signals are evolving both in time and in space. Their observations represent multivariate time-series of samples acquired uniformly in time, but often non-uniformly in space. It is claimed that temporal and spatial domains of multi-dimensional signals are equivalent in computing the signal transformations and metrics. The aim of this paper is to define space-time correlations of multivariate time-series data generated at multiple spatially dispersed sensors. In order to reduce a possibly large dimensionality of the spatial index, it is proposed to assume a one-dimensional spatial shift along a defined low-dimensional spatial trajectory or manifold. Moreover, the sample alignment between the space-time shifted multi-dimensional signals can be defined as the nearest neighbor problem. A public EEG dataset is used as an example to calculate the intra- and inter-cluster space-time correlations and also distributions of signal correlations versus the sensor distances.