Indoor positioning technology based on the Wireless Local Area Network (WLAN) fingerprinting method is becoming a promising choice as for ubiquitous WLAN infrastructure. The technology mainly compares the received signal strength (RSS) of a mobile device with an RSS fingerprint in the fingerprint database, and uses the matching rule to find the closest match as the estimated position of the device. The quality of the fingerprint database construction can directly affect the positioning results. This work proposes a three-stage fingerprint database processing method. In the first stage, the original fingerprint database is divided into several small sub-fingerprint databases according to the specified rules. In the second stage, every sub-fingerprint database is processed using the principal component analysis method to achieve a reduced dimension fingerprint dataset. In the third stage, the k-d tree method is used to process each dimension-reduced sub-fingerprint database for obtaining a hierarchical sub-fingerprint database. In addition, in the online phase, the best bin first (BBF) method is applied to the search engine of sub-fingerprint database to complete the location determination of the device. This method can improve positioning performance through simulation research.