Online Motion Mode Recognition for Portable Navigation Using Low-Cost Sensors

The main challenge in portable navigation is obtaining accurate positioning results in urban canyons and indoors where global navigation satellite system signals are degraded/unavailable. Inertial sensors can be used; however, their traditional positioning performance is unacceptable. Motion mode recognition of the user carrying the portable device is needed to enhance the positioning performance. This paper illustrates the use of pattern recognition to detect stationary (including standing, sitting, and placing device on still surface), walking, running, cycling, and vehicle (including car, bus, and train). The use of sensor fusion to remove bias errors from sensor signals to calculate variables descriptive of motion has been introduced in this paper. The features are extracted from such variables and combined into a feature vector fed into the classification stage. Extensive data collection was conducted to gather training and evaluation data; the performance results on the latter have shown the efficacy of the solution. Copyright (C) 2015 Institute of Navigation