This master thesis deals with the problematic of the integration of Global Positioning System (GPS) measurements with inertial data acquired by a Micro-Electro-Mechanical System (MEMS) Inertial Measurement Units (IMU). This technology is employed for the purposes of sport performances assessment, as it enables reconstructing accurately athletes trajectories. Based on a recent development at the TOPO laboratory at the Swiss Federal Institute of Technology of Lausanne (EPFL), a new software's architecture is proposed to ensure an automated treatment of the input data. In its first phase, a Continuous Wavelet Transform (CWT) is performed to split the signal as a function of its dynamic. Then, quasi static periods are automatically identified to initialize the processing. Thereafter, several ranges can be integrated in order to compute an optimal trajectory. The performances of this new architecture are validated and evaluated using several sport experiments in skiing and biking. The implemented method is reliable and works correctly. The process offers the capability of bridging lacks of GPS data lasting up to 10 seconds, without any substantial degradation of the trajectory's accuracy. In the frame of this project, a new MEMS-IMU was also engaged, in order to evaluate its navigation performances. It appears that its stochastic model needs to be refined and a specific initialization strategy developed before this sensor finds its place in trajectory reconstruction for downhill skiing.