In this work, methods to detect one or several change points in multivariate time series are reviewed. They include retrospective (off-line) procedure such as maximum likelihood estimation, regression, kernel methods, etc. In this large area of research, applications are numerous and diverse; many different models and operational constraints (on precision, complexity,…) exist. A formal framework for change point detection is introduced to give sens to this significant body of work. Precisely, all methods are described as a collection of three elements: a cost function, a search method and a constraint on the number of changes to detect. For a given method, we detail the assumed signal model, the associated algorithm, theoretical guarantees (if any) and the application domain. This approach is intended to facilitate prototyping of change point detection methods: for a given segmentation task, one can appropriately choose among the described elements to design an algorithm. A review of change point detection methods