TSSummarize: A Visual Strategy to Summarize Biosignals

Abstract

Visual tools enhance the human ability to detect structures found on time series. Medical doctors and data-scientists rely on their visual abilities to perform time series analysis. A visual tool that would summarize several sources of information of time series would be of great value and is not yet provided in the literature. This work proposes a novel unsupervised visual strategy to summarize a time series and compact several layers of information. The strategy extracts information from the Self-Similarity Matrix (SSM). This data source is able to segment the time series, detect events and show relationships between subsequences. The visual strategy has been tested on several use-cases from the medical domain, proving to be type agnostic, intuitive and compact.