| The people’s life has undergone tremendous changes as the development ofscience and technology. And positioning technology as a key demand both in thecommercial or military has been more and more significance in everyone’s life,work, security and so on.This article get the idea of information fusion based on large data processing, itis freed from the causal relationship between the pursuits and to focus on therelationship between discovery and application. We begin to pay attention to largechanges in the overall and action signal rather than focus on accuracy of local signal.So that we can overcome the problem of accuracy, the disturbance of noise and thegyroscope drift. At the same time, we plan to use the technology of intelligent actiondecision based on support vector machine. According to the signal characteristicsdetermine the action. Combined with the individual attribute information such as age,gender, stride length of walk, run, down stairs and up stairs to get the staff locationin real time. This is the main contents of this paper is divided into the followingareas:First of all, the core algorithm of wavelet packet and the system of supportvector machine algorithm is introduced. And some simple mathematical derivationis showed. We have some theoretical research in frequency and time domain, andpropose treatment schemes, laid the theoretical foundation for the application of thealgorithm.Secondly, we study the characteristics and model of pedestrian walking andanalysis the physical process of personnel walking stages. We get the best positionto wear sensor device, and analyzed the data waveform when personnel walking.Changes derived from the moving target action are complex and the output signalsof motion sensor are typical non-stationary signal. This paper studies the vibrationsignal for processing algorithms, including the time domain analysis algorithm andthe frequency domain analysis. By comparing the characteristics of their functions, we finally determined to use the method of wavelet decomposition of the signalfeature extraction of different values for action. Because walking, running, up anddown the stairs will cause the change of vibration energy, different actions will causeenergy changes in different frequency band. We use wavelet packet analysis methodbased on calculation of different frequency band to extract the feature.Thirdly, in order to improve the efficiency and accuracy. The paper studiedSVM support vector machine in this aspect. SVM as the effect of learningclassification method training is very good and popular. This system combined themotion characteristics of wavelet extraction method of these feature values into theSVM training and decision can implemented action intelligent decision.Finally, we design the signal acquisition system of gyroscope, accelerometerand the electronic compass three sensors signal processing. We complete the dataacquisition equipment and the host computer of the signal processing. We completea set of online detection and judgment of the indoor location system both onhardware and PC software development. As hardware for data acquisition, we usethe microcontroller of Cortex-M3core. Software development using virtualinstrument of LabVIEW based on the concept of programming language to programa good human-computer interaction interface to complete the vibration signal datatransmission, processing, storages of a series of functions, algorithm and locationwith a stronger practicality. According to the characteristics and model of pedestrianwalking and clarified the personal attribute classification of gait, the step numberestimation, stride width estimation, and stride heading estimation, using the outputvectors. We can realize the location of person on two degree plane. |
PDF Full Text Request