*1.2. Contribution*

Most of the described algorithms were evaluated either for walking or for running on a treadmill. However, both of these conditions yield different signal characteristics to those of free running. In running, different strike patterns, such as rearfoot or forefoot strike, exist and affect the performance of these algorithms. Besides, the movement is also more dynamic, which yields higher accelerations, angular rates, and impacts. Therefore, our contribution is the comparison of different algorithmic approaches for computing average velocity and stride length during overground running using an IMU embedded into the sole of a running shoe. We evaluate these algorithms on a large database including high variation of the input data. Additionally, we run a field study to assess the performance in a real-world scenario. Based on the results, we give implementation recommendations for specific use cases.
