Distance measures play a crucial role in quantifying the similarity and dissimilarity between two intuitionistic fuzzy sets. Numerous distance measures have been established in prior research, but some exhibit limitations in effectively differentiating intuitionistic fuzzy sets from high hesitancy degrees. This paper proposes an improved distance measure and further demonstrates its efficacy in pattern recognition applications. Firstly, through systematic case analyses, we identify two critical shortcomings in existing distance approaches, i.e., 1) insufficient capture of cross information between membership degrees and non-membership degrees, and 2) inadequate handling of high hesitancy scenarios. Then, an improved distance measure is proposed by comprehensively considering membership degrees, non-membership degrees and high hesitancy degrees. Furthermore by the improved distance measure, a pattern recognition algorithm is proposed. Finally, the new approach of pattern recognition is concretely applied in two practical examples of medical diagnosis and road bidding, so that the relevant effectiveness is verified.