Human movement recognition method based on cross-domain dictionary learning

Abstract

The invention provides a human movement recognition method based on cross-domain dictionary learning. The method comprises the steps that intra-class differences between training samples of a target domain are expanded through a cross-domain movement recognition framework based on a source domain data method so as to enhance the performance of an existing recognition system, and therefore annotated data obtained from a different domain are used as a source domain to be provided on the basis of manually-annotated movement information in a target domain; a domain self-adapting dictionary pair is learned through a discriminating cross-domain dictionary learning method, and data with different distributions are migrated to the same feature space so as to enable feature distribution of different domain data to be matched; corresponding annotated information between different domains is not executed so as to adapt to different kinds of migration learning in reality. According to the method, data distribution of relevant movement of a source domain database is used for adapting to data distribution of movement in the target domain, and the dictionary pair with reestablishment, discrimination and domain self-adaptability can be learned and obtained through the method.

Description

technical field [0001] The invention relates to the technical field of computer vision tasks, in particular to a human action recognition method based on cross-domain dictionary learning. Background technique [0002] Recently, dictionary learning for sparse representations has attracted a lot of attention. It has been successfully applied to many computer vision tasks, such as face recognition (see "J.Wright, A.Y. Yang, A. Ganesh, S.S. Sastry, and Y. Ma. Robust face recognition via sparse representation. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 31(2):210–227, 2009.") and image noise reduction (see literature "M.Zhou, H.Chen, J.Paisley, L.Ren, G.Sapiro, and L. Carin. Non-parametric Bayesian dictionary learning for sparse image representations. 2009.”). Using an overcomplete dictionary, the sparse signal model is able to estimate the input signal through a sparse linear representation of the dictionary basis. According to different criteria, many me...

AI Technical Summary

Problems solved by technology

In these cases, it is expensive or even impossible to obtain more training samples

Images