A Similarity Measure Between Vector Sequences with Application to Handwritten Word Image Retrieval

This article proposes a novel similarity measure between vector sequences. Recently, a model-based approach was introduced to address this issue. It consists in modeling each sequence with a continuousHiddenMarkovModel (CHMM) and computing a probabilistic measure of similarity between C-HMMs. In this paper we propose to model sequences with semi-continuous HMMs (SC-HMMs): the Gaussians of the SC-HMMs are constrained to belong to a shared pool of Gaussians. This constraint provides two major benefits. First, the a priori information contained in the common set of Gaussians leads to a more accurate estimate of the HMM parameters. Second, the computation of a probabilistic similarity between two SC-HMMs can be simplified to a Dynamic Time Warping (DTW) between their mixture weight vectors, which reduces significantly the computational cost. Experimental results on a handwritten word retrieval task show that the proposed similarity outperforms the traditional DTW between t...

• Type: Conference
• Year: 2009
• Where: CVPR
• Authors: José A. Rodríguez-Serrano, Florent Perronnin, Josep Lladós, Gemma Sánchez
• Keywords: Computer Vision, CVPR 2009, Handwritten Word Image Retrieval, Similarity Measure