Bayesian topological signal processing (Q2129149)

This paper works on giving an interpretable framework for signal processing via sublevel homology. The paper is trying to establish interpretable links between sublevel set persistence diagrams of signals and their frequency domain which are used to study time series analysis. The authors explore Bayesian framework for the time series classification, from which they find that the Bayesian topological features are pretty useful just as other well-known features, such as those from power spectral densities and continuous wavelets. Finally, they apply their results to electroencephalography which is a neuroimaging technique wherein electrodes are placed on a subject's head to measure local changes in voltage over time, which are reported as a collection of time series.