Most of the day to day people activities are carried out inside buildings. Many sectors such as, medicine, industry, academia or even security systems require indoor positioning services. As a consequence, it is essential to develop a reliable and accurate indoor positioning system (IPS). Since global navigation satellite systems (GNSSs) are not suitable for indoor localization, several IPSs have emerged. However, each indoor positioning technology has its advantages and disadvantages. Hence, there is not an IPS system with the best performance for every situation.

The IPS databases based on the Wi-Fi infrastructure installed in two buildings of the Tampere University of Technology required an update. Therefore, the scope of this thesis has been to update and moreover, optimize the IPS fingerprint databases of these two buildings. The results have been presented and analyzed with the expectance that they will be useful for similar or wider projects.

Multiple IPSs are explained, as it is convenient to understand the advantages and the weaknesses of each technology. The technology which provides the positioning services is the fingerprint Wi-Fi received signal strength (RSS). In that way, a measurement database is built. The database is used to simulate the IPS, which is implemented through the Bayesian estimation algorithm and the k-nearest neighbors technique. Successively, the parameters of the algorithm are optimized.

The analysis of the results showed that for the lowest values of the parameters, the performance of the system improves with respect to higher values of the parameters. The best performance of the Wi-Fi based IPS results in a floor detection probability nearby 99% and an average distance error below 3 m. However, negative effects, such as the ones produced by outlier measurements, must be taken into account. Some weaknesses of the Wi-Fi based IPS, such as the challenges associated to the training phase, open a path of research that might enhance the system performance.