This thesis introduces an interface for semi-automatic labeling of mobile sensor data which integrates automatic data clustering with the user expertise. This method saves a lot of time and eort because the amount of data to be manually labeled is decreased since the user does not have to label individual samples but instead groups and clusters of data. This method not only solves the problem of labeling but also allows user-specic recognition systems rather than the systems trained to recognize a specic number of activities. The main goal for this thesis is to develop a graphical user interface for labeling mobile sensor data which helps in the training of motion recognition systems. An initial evaluation of the application showed that the pro- posed solution could be an appropriate solution for the labeling problem. There are a lot of open topics regarding the labeling problem that could be an interesting target for future work where adjustments or extensions could be done for this project in order to reach an ecient solution.