Research On Gait Recognition Based On Tactile Characteristics Of Plantar

Along with the development of biometric recognition technology, the biometric such as fingerprint and iris have been widely used in the field of identification, but these biometric all have some limitations, gait recognition with the unique capability which can recognize people at long distance, difficult to disguise, imitate and noninvasive, has been getting much attention. Nowadays the gait recognition methods are mostly based on visual, the image which through these methods is often fuzzy due to the shot distance, light and the movement of people, also the climate, scene, clothing and other interference maybe influence on image capture. So this thesis proposes a new gait recognition method based on plantar tactile characteristics, the main work and research results are as follows:1. Collection the static and dynamic tactile gait data by the designed procedure. This thesis adopt the pressure test board designed by intelligent machinery research institute of Chinese academy of sciences to collect the gait data which conclude51individual’s both feet data, then save the right and left foot’s data separately. To the static, save the data which people stand stability, to the dynamic, save the data which everyone’s each foot walk through the board10times at normal, quick and slow speed, the below process are all getted dowm to the right and left foot’s data separately, also training and test the data of left and right foot separately.2. Propose the key frames extraction and plantar subdivision method. First, extract key frames as sample data, to the static data, extract16frames as key frames from everyone’s30frames data, to the dynamic data, extract the data which sum pressure is largest, and the preframe, the next frame of this data as key frames from every step walking, then take the process of area subdivision, divid the palntar into six regions which according to the method this thesis proposed.3. Extract the kinetic characteristics, Laplace spectrum characteristics and shape characteristics which used to training and test. The first part is the kinetic characteristics, including the plantar maximum pressure points, center pressure points and the pressure ratio; The second part is the Laplace spectrum characteristics based on maximum pressure points and center pressure points; The third part is the foot shape characteristics according to-the shape of the plantar, integrate the three parts feature is the feature vector which this thesis construct.4. Design the static and dynamic classifiers. This thesis select one-to-one SVM (Support Vector Machine, SVM) classification method to construct classifiers, for static gait recognition, let everyone’s static sample data as one class, for dynamic gait recognition, aimed at different speed of gait data, design three kinds of classifiers, namely independent classifiers, simple classifiers and mixture classifiers.5. Make the static and dynamic gait recognition experiment. In static gait recognition, we need to comprehensive the recognition result of left and right foot to get the finally recognition result, in the dynamic gait recognition, the left and right feet’s accuracy rate are calculated respectively by three classifiers. The experiment results show that the proposed gait recognition method based on plantar tactile characteristics is simple on operation, interference has little influence on it, not only can apply to static gait recognition, but also can apply to dynamic gait recognition which used independent classifiers, achieved a high recognition accuracy.