Improving GPS-based indoor-outdoor detection with moving direction information from smartphone

This paper presents an improvement for GPS-based indoor-outdoor detection with moving direction information. First, we considered several features based on raw GPS satellite informationas the baseline. From the evaluation with GPS data from 19 persons, the SVM-based classifier using a combination of elevation and S/N ratio achieved over 96% accuracy for the simple 'clear' situation. Then we introduced direction information from a compass sensor to increase the detection robustness in the canyon of buildings. The second experimentation result showed the proposed method considering direction information kept almost the same accuracy of indoor-outdoor detectionin a 'canyon' situation, whereas the baseline classifier worsened accuracy by 10%.