Detection of quasi-static instants from handheld MEMS devices

In this paper, an algorithm for the detection of quasi-static instants (QS) from handheld MEMS devices is presented. In order to tune the detector according to the variety of motions that the hand can perform, a decision tree classifier, able to recognize activities typical for mobile phone users, such as phoning, texting, walking with swinging hand or carrying the device in a bag, has been designed and implemented. Performances of the proposed detector of QS epochs and of the motion mode classifier are assessed with experimental data collected with several individuals. In addition, the relationship between QS instants and human gait is investigated. Specifically, the use of QS instants for the detection of the user´s step is analyzed.