In order to solve the problem of large error and low efficiency when the WiFi electromagnetic fingerprint localization method is used to distinguish floors in indoor floor identification. An indoor floor recognition method based on improved hierarchical clustering（ HC） algorithm is presented. The hierarchical clustering algorithm is improved by density stratification. This method solves the problem that the hierarchical clustering algorithm is sensitive to isolated points

and improves the accuracy of indoor WiFi fingerprint data clustering.Moreover

the fingerprint data in the off-line stage are cut and the larger datasets are transformed into clusters of small datasets

which reduces the computational complexity of the fingerprint location data. The simulation results show that the proposed floor recognition method can effectively improves the positioning accuracy. Compared with hierarchical clustering algorithms such as K-means

CURE

ROCK

this method has better localization effect in floor discrimination.