A hybrid classification approach to improving location accuracy in a Bluetooth-based room localisation system

It has been well recognised that the use of localisation techniques in home environment are beneficial to the development of health monitoring and activity recognition systems. The Bluetooth devices, as a kind of effective sensor with remarkable characteristics such as low cost, have been widely used in our daily life. Research has been carried out to integrate cellular network signal measurements and Bluetooth link measurements in developing home localisation systems. This paper presents a hybrid classification approach, based on the combination of Bayesian statistics and supported vector machines, to supporting the development of the Bluetooth-based room localisation system. The proposed approach considers the dependency between features and non-linear overlapping of features between rooms. The results show that the prediction accuracy has been improved in comparison to the traditional Naive Bayes classifier and the hidden Markov model used in previous studies.