| Nowadays, GPS position system is a kind of mature navigation system, which the accurate of in outdoor can meet majority of civil and military needs. When the GPS signal goes through the walls, it becomes so weak that terminals can’t be positioned or positioned with high price. With the continuously development of technology, people spend more and more time on indoor activities. Indoor location technologies are needed. There is not a mature and universal location system in several kinds of technologies until now. Indoor location technologies, such as WiFi position technology and L-band wireless position technology, have a lot of problems, for example, multipath, refraction, reflection, signal instability. In addition, inertial sensors location technology is widely used in indoor and outdoor position with its virtue of autonomous navigation, sustained position, no dead corner etc. There are still two issues. Firstly, the majority of intelligent terminals use consumer sensors considering of costing. Because of consumer sensor’s not high accuracy, it will lead to low accuracy of inertial sensors location. Secondly, according to human habits, intelligent terminals generally hold by hand or hands, thus the accuracy of sensors location will be affected by human’s specific motions. Targeting at the above existing tissues, this dissertation has the main contributions as follows:1. FIR Low-pass Filter designed with Hamming window’s size self-adjusting based on stride-frequency.After tested by practice, human’s walking stride-frequency belongs to low frequency range. In order of getting rid of noises and interference, this dissertation proceed a(t) with FFT transform, so as to determine human’s walking stride-frequency range, and then set the size of Hamming window according to the range, and finally apply it to execute low-pass filtering. Experimental results show that this filtering algorithm has achieved anticipated filtering result because it can not only remove ambient noise, but also can maintain characteristics of a(t)-the signal well.2. Self-learning-based step-length estimation algorithmIn Pedestrian Dead Reckoning algorithm, step-length influences reckoning accuracy the most. It will highly improve the accuracy of sensor location technology, if step-length is estimated accurately. In this dissertation, algorithm divides into two parts:learning of standard characteristic offline and step-length level identification online. Learn the characteristics of different levels of step-length in the offline, while comparing human’s walking characteristics the real-time to different levels of step-length, identify which to match, so as to estimated step-length.3. Magnetic compass error compensation based on multi-sensors systemMagnetic compass determines orientation of sensor according to component of the geomagnetic. In daily use, the accuracy of orientation, which is easily interfered by the magnetic field round, results in lower accuracy in pedestrian dead reckoning algorithm. Targeting at the issue, this dissertation has put forward a method to error compensation, such as soft and hard magnetic correction, tilt compensation, magnetic declination compensation, based on multi-sensors. In soft and hard magnetic correction, the dissertation introduces ellipsoid model. In the tilt compensation, the dissertation introduces gyroscope as the reference platform, further improve the accuracy of orientation.4. Specific walking behaviors identification of human bodyNormally, people will move forward at a constant speed and this is a default behavior in the pedestrian dead reckoning algorithm. However, as the self-conscious creatures, people walk flexibly. So research of default behavior is bound to encounter bottleneck problem in the practical applications. Targeting at the issue, this dissertation mainly focuses on the identification of the specific behavior of the human body--forward, backward walking identify and horizontal walking behaviors recognition. |
PDF Full Text Request