Learning Temporal Sequences using Bags of Features

This paper presents a novel method for learning classes of temporal sequences using a bag-of-features approach. We define a temporal sequence as a bag of temporal features and show how this representation can be used for the recognition and segmentation of temporal events. A codebook of temporal descriptors, representing the local temporal texture, is automatically constructed from a set of sample sequences at multiple time scales. Temporal sequences are then encoded using accumulated histograms of parts from this codebook. This representation, though simple, proves to be surprisingly powerful and able to implicitly learn the sequence dynamics. Based on this representation, a multiclass classifier, treating the bag of features as the feature vector, is applied to estimate the corresponding class of the temporal sequence. Finally, extensive experiments are performed on multiple datasets to compare our method against state-of-the-art algorithms. The results show that our algorithm performs better and requires less training data than competing techniques.