Positioning is a crucial task in pervasive computing, aimed at estimating the user's positions to provide location-based services. In this paper, we study an interesting problem: when we wish to obtain hybrid positioning granularities in an office environment, how can we incorporate heterogeneous sensors to build an indoor positioning system with minimal human calibration effort? We propose a calibration-less solution by incorporating the ultrasound sensors with the radio-frequency sensors. In our solution, we use these two different types of sensors to satisfy the different granularities requirement; and meanwhile, we use the ultrasound sensors to help calibrate the radio-frequency sensors for positioning, so that we can minimize, or even eliminate the labeling effort for the radio-frequency positioning. Finally, we develop a system prototype with real-world sensor networks, and verify the feasibility and effectiveness of our proposed solution.