Indoor positioning systems (IPSs) are attracting more and more attention from the academia and industry recently. Among them, approaches based on WiFi techniques are more favorable since they are built upon the WiFi infrastructures available in most indoor spaces. However, due to the bandwidth limit in mainstream WiFi systems, the IPS leveraging WiFi can hardly achieve centimeter localization accuracy under strong nonline-of-sight (NLOS) conditions which is common for indoor environment. In this paper, to achieve the centimeter-level accuracy, we present a WiFi-based IPS that exploits the frequency diversity via frequency hopping. In the offline phase, the system collects channel frequency responses (CFRs) from multiple channels and from a number of locations-of-interest. Then, the CFRs are post-processed to mitigate the synchronization errors as well as interference from other WiFi networks. Then, using bandwidth concatenation, the CFRs from multiple channels are combined into location fingerprints which are stored into a local database. During the online phase, CFRs are formulated into the location fingerprint and is compared against the fingerprints in the database via the time-reversal resonating strength (TRRS). Finally, the IPS determines the location according to the TRRS. Extensive experiment results demonstrate a perfect centimeter-level accuracy in an office environment with strong NLOS using only one pair of single-antenna WiFi devices.