Sliding windows and persistence: an application of topological methods to signal analysis (Q2355335)

This paper develops a theoretical framework for the topological study of time series data. A particular interest is the detection of periodicity. The approach is to to study the persistent homology of sliding window embeddings. This seems quite natural as persistent homology is based on noting the birth and death of homology classes. In this case the time axis provides the level map required for persistent homology. Periodicity can be inferred in 1D persistence from the maximum persistence based on window size which measures the \textit{roundness} of the data cloud. The authors motivate and illustrate their approach by examining truly periodic examples, which is a great help for the reader. The methodology is applied to identify periodicity in synthetic data sets and compared with those used in state-of-the-art methods in gene expression analysis.