Research Of Gait Recognition And Gait Symmetry Based On Machine Vision

Walking is a extremely common behavior in people’s everyday life.The characteristics presented when human walking has highly important value.Gait analysis that studys on characteristics of the human body has been applied to many fields,such as biological dynamics,clinical medicine,humanoid robots,rehabilitation robots, biological recognition and so on.Using skeletal data captured by Kinect sensor,which is simple to use,small and very cheap,this thesis studies two content consist of gait recognition and gait symmetry.First of all,this thesis recognizes different persons in the same road conditions.The 20 3-D skeletal points obtained by Kinect is used to construct covariance matrix in order to indicate the contract between these points.Using K nearest neighbor classifier,the gait is recognized by computing the similarity of covariance matrix between the testing data and the training data.Average recognition accuracy of 89% has been achieved for a experiment consisting of two walking routes of 11 subjects,which shows that the proposed method is robust to variations such as the walking direction.Then,taking advantage of PCA reconstruction algorithm,this thesis uses motion data of the left lower limb to reconstruct the motion data of right lower limb.Compareing to the original information of the right lower limb,a conclusion is drawn that the lower limb gait of healthy human is symmetrical.On this basis,in accordance with the problem that the original gait pattern of the hemiplegia and amputations patient has most probably not been recorded,this thesis proposes a average model using healthy people’s average gait information to construct the patient’s and demonstrates the rationality of the model.Finally,this paper proposes a symbol-based approach to quatify the gait symmetry.The data is segmented according to a piecewise linear segmentation algorithm so as to preserve its dynamic information and overall shape.After feature extraction and clustering,the segments is symboled according to its cluster.On the basic of period histogram,symbol-based symmetry index is used to analysis a data set including normal space walking,slow space walking and intentional asymmetry.Gait analysis can convey important information about one’s physical and cognitive condition.This thesis focuses on two aspects of gait recognition and gait symmetry. According to the proposed approach,this thesis designs some experiments to capture skeletal data.After discussion and analysis in detail,the proposed model is proved to be rational and reliable.In addition,pretty good results have been achieved.