01525nas a2200265   4500000000100000000000100001008004100002260001200043653001700055653001300072653002100085653001500106653002300121100001700144700002100161700002000182700001600202700002300218245007900241856009800320300001000418490000600428520081100434022001401245       2015                            d  c12/201510aLocalization10aLearning10aMachine Learning10aAlgorithms10aInformation Fusion1 aPaulo Novais1 aRicardo Anacleto1 aLino Figueiredo1 aAna Almeida1 aAntónio Meireles00aStep Characterization using Sensor Information Fusion and Machine Learning  uhttp://www.ijimai.org/journal/sites/default/files/files/2015/11/ijimai20153_5_7_pdf_23101.pdf  a53-600 v33 aA pedestrian inertial navigation system is typically used to suppress the Global Navigation Satellite System limitation to track persons in indoor or in dense environments. However, low- cost inertial systems provide huge location estimation errors due to sensors and pedestrian dead reckoning inherent characteristics. To suppress some of these errors we propose a system that uses two inertial measurement units spread in person’s body, which measurements are aggregated using learning algorithms that learn the gait behaviors. In this work we present our results on using different machine learning algorithms which are used to characterize the step according to its direction and length. This characterization is then used to adapt the navigation algorithm according to the performed classifications.  a1989-1660