The complexity and diversity of indoor environments brings significant challenges to automatic generation of navigation instructions for blind and visually impaired users. Unlike generation of navigation instructions for robots, we need to take into account the blind users wayfinding ability. In this paper we introduce a user-centric graph based solution for cane users that takes into account the blind users cognitive ability as well as the user's mobility patterns. We introduce the principles of generating the graph and the algorithm used to automatically generate the navigation instructions using this graph. We successfully tested the efficiency of the instruction generation algorithm, the correctness of the generated paths, and the quality of the navigation instructions. Blindfolded sighted users were successful in navigating through a three-story building.