Gait recognition on smartphones could be considered as one of the most user-friendly biometric modalities. The main benefit of gait recognition is that it is an unobtrusive biometric modality, since it requires little interaction with the user. Users would only have to carry the sensor device and walk as normally. Its unobtrusiveness make it suitable for a user-friendly access system. Up to date, most studies on gait recognition have been done using dedicated hardware acquisition sensors. Nevertheless, one possible solution for gait recognition is using sensors embedded on smartphones. This paper compares the performance of four state-of-art algorithms on a smartphone. These algorithms have already been tested on dedicated hardware but not in a commercial phone. For such purpose, a database using a smartphone as acquisition device has been obtained. State-of-art gait recognition algorithms have been tested on this data base, as well as a new cycle detection algorithm which has been designed to have the same starting point. As a result, the algorithms have shown EER ranging from 16.38% to 29.07%, These EERs are significantly higher than the ones obtained in dedicated hardware which ranged from 5.7% to 13%.