This paper investigates different strategies allowing integration of contextual information during the feature extraction stage of a cursive handwriting HMM-based recognition system. First we propose to use linear discriminant analysis (LDA) in order to integrate the class information during feature set building. Secondly several zoning strategies are used to integrate local contextual information. Finally, a weighting technique is proposed in association with zoning with the aim of integrating handwriting style. Some experiments were carried out and the results show the interest of the proposed strategies.