Research On Acceleration Signal Based Recognition Of The Human Upper Limb Motion

In the field of human-computer interaction, the technology ofnatural human-computer interaction is gradually becoming as a newdeveloping direction of human-computer interaction. It is in pursuit of newway of human-computer interaction towards natural, harmonious, andhumanized. As a kind of natural way of human-computer interaction, thetechnology of human upper limb motion recognition based on accelerationsignal has widely attracted the attention of researchers, and shows greatpotential applications in lots of fields, such as the intelligenthuman-computer interaction、 virtual reality、 body feeling game、 medicalrehabilitation training。The human upper limb motion recognition based on acceleration signalis a new kind of motion recognition technology. Although the technology hasachieved rapid development in recent years, it is still in the relative basicresearch stage. To carry on the research on motion recognition, the processusually involves three steps: motion signal acquisition, motion dataprocessing and motion pattern recognition and prediction. There are threemain problems to consider during the research work of the upper limb motionpattern recognition based on the acceleration signal, that are how to designthe acceleration signal acquisition scheme of the upper limb motion, how tocarry on signal obtained processing, and how to achieve the high precisionmotion pattern recognition. This topic mainly discusses these issues, andlaunches a series of studies. In this dissertation, the main research work doneis listed as follows:Firstly, on the basis of consulting a large number of domestic andforeign literatures, the research background and significance of this topic arebriefly summarized, and analysis the research situation in related fields athome and abroad. Determine the specific research content, and the general research scheme is put forward.Secondly, a set of data acquisition device is designed and developed. Anew acceleration signal acquisition method of the human upper limb motionis put forward. The3-axis acceleration sensor ADXL345is selected, and hasdeveloped a set of motion acceleration signal acquisition device. There aretwo aspects about the acquisition device are introduced respectively, that iscircuit design of its hardware system and the design of software acquisitionprogram.Thirdly, combined with wavelet theory, a kind of upper limb motionrecognition algorithm based on statistical pattern recognition method is putforward. The preprocessing methods including the wavelet thresholdde-noising method and normalization are introduced. For thethree-dimensional acceleration signal after pretreatment, extractingtime-domain feature values, energy information values in the threedimensions with wavelet packet decomposition as while, to build the featurevector commonly. At the last, some relevant theories about the SVMclassifier and the open source software LIBSVM for SVM classificationalgorithm are introduced respectively.Finally, the specific experiment is designed to do validation and analysison the general research scheme. To carry out the experiment,10test subjectsare selected to involve in data collection, and seven kinds of upper limbmotion patterns are defined. To obtain the original data set for the researchexperiment, collecting30times of each motion pattern for every experimentobject. Since the original data set cannot be directly used for classificationrecognition, it is required to do preprocessing and extraction of characteristicparameters. The much better generalization capability of the SVM classifieris built based on the extracted feature vector. In order to get the optimaldecision model of SVM classification, the optimal RBF kernel function isdetermined by comparing the experimental results of four different kernelfunctions for SVM. The optimal kernel parameters (C,) is obtained by gridsearch algorithm. At the last, the cross-validation experiments are carried outon two different types of data sets. The experiment results are analyzed and discussed. And then the research plan of this topic is verified to be effective.