Translated Abstract
The indoor inertial navigation is based on the pedestrian dead reckoning (PDR) algorithm and includes three parts: step number, step length and walking direction. In the process of indoor PDR location for pedestrians using smart phones, the mobile phone often needs to be placed in the hand stably meanwhile the top of which must be placed forward. However, in the actual application, a mobile phone may be in a pocket, a backpack, a shaking hand, which makes it difficult to detect steps accurately and to obtain the direction of walking directly using the electronic compass. In order to overcome the decline in the detection precision of the step number in the free posture of the mobile phone and the failure of the estimation of the moving direction, starting from the attitude algorithm, we use the idea of the finite-state machine (FSM) and the statistical method of the principal component analysis (PCA) to track the trajectory of pedestrians when the attitude of the mobile phone is not determined.
We propose a multimodal pedestrian dead reckoning (MPDR) algorithm based on attitude algorithm in this paper for the the mobile phone is relatively stable to the human body while the the attitude of the mobile phone is not determined, such as putting the phone in the pocket or backpack. There are three aspects of the algorithm. First, the FSM step detection method for multi-attitude is improved and the thresholds is updated adaptively to ensure the accuracy of the steps. The multi-mode FSM step number detection method has obtained 98.41% average detection rate in the hand, pocket and backpack, which has been improved by 14.18% compared with the FSM step detection method. Then a method of fuzzy judgment whose accuracy of the judgment up to 99.41% is proposed for the statistics of the horizontal acceleration, so that the final direction is obtained from the result of PCA processing. Finally, the fuzzy judgment confidence is defined, and the walking direction is corrected by confidence level and gyroscope.
We propose a swing-mode pedestrian dead reckoning (SPDR) algorithm based on attitude algorithm for the mobile phone moves forward and backward in hand along the walking direction in this paper. This algorithm consists of three aspects. First, a peak detection step method based on horizontal acceleration is proposed. The step accuracy rate of 97.65% is obtained under the condition of obvious and regular swing movement. Then the corresponding finite-state machine is designed to carry out step detection, and a method to updating thresholds adaptively is proposed to judge the states. The average detction rate of 98.77% is obtained in general circumstances. Finally, a direction estimation method is introduced, which combines the PCA technology and the compass angle. This part also involves a method of directional quantization, which can improve performance at some special time.
Translated Keyword
[Attitude analysis, Finite-state machine, Pedestrian dead reckoning, Principal component analysis]
Corresponding authors email
