Bout categorization and motif segmentation of the dataset from Marques et al.

(a) The five bout types identified using a Gaussian Mixture Model. Each bout is characterized using its displacement, change in heading and the first four PC components of the tail angles. (b) The histograms of displacement and change in heading for the five bout types. (c) A sample from the dataset showing the most likely segmentation of the data into motifs from the learned dictionary.

(EPS)