Description of Invention
In indoor positioning, a large Bluetooth low energy (BLE) fingerprint database is often constructed as a result of a multi-floor indoor environment commonly observed in the real world scenario. Consequently, this leads to a high computational complexity and increased computational time when using only the baseline localization algorithm during location prediction. To overcome this issue, a clustering-based indoor positioning system (IPS) known as DECIPS is proposed to reduce the computational complexity and execution time required by the localization algorithm for location prediction. Deep embedded clustering (DEC) algorithm is adopted in the proposed DECIPS to separate the large dataset into subsets before using them to train classifiers and regressors established specifically for each cluster. Thereafter, the performance of DECIPS is compared with several clustering-based IPSs in terms of their average positioning error and execution time.