Harmony search (HS) is a meta-heuristic algorithm imitating the music improvisation process, which has been successfully applied to many real-world problems. This paper presents a harmony search algorithm based on dynamic behavior selection (DBSHS) for solving high dimensional function optimization problems. Improvisation of a new harmony has three strategies in the DBSHS. In the process of iteration, the improvising behavior of the harmony is determined by calculating the immediate value and the comprehensive value of each strategy, individual improvising strategy selection method is proposed to improve the individual search speed or to avoid falling into the local optimal solution. The DBSHS is compared with nine variants of harmony search on 22 benchmark functions. The experimental results show that the proposed DBSHS has good solution accuracy, remarkable stability and high convergence speed. It is particularly good at solving complex high-dimensional problems.