In this paper, we propose a GPS-less localization method, called Hidden Access Point Estimation-based Localization or HAPEL to pinpoint user's current position in an underground or indoor environment, especially where APs are placed scarcely in that two or less Wi-Fi access points are within the scanning range. Conceptually, HAPEL enhances the weighted centroid localization by exploiting hidden access points and estimating the direction of user movement. The movement direction can be estimated with digital compass, accelerometer and gyroscope all of which are usually installed on smartphones, and the estimation result is used to choose which hidden access points participate in the weighted centroid localization. We have conducted a performance evaluation study on HAPEL in an empirical test-bed. The results indicate that HAPEL improves the accuracy of the weighted centroid localization significantly.