Proposing a new set of features based on frieze pattern to discriminate normal and abnormal gait

In this article a markless vision based technique for extracting a set of features from human walking sequence for differentiating normal and abnormal gait is presented. Gait change is an indicator in many diseases. Clinical gait assessment tools were proposed to asses ones gait but the result of these tools depend on raters´ opinion. Consequently, quantitative gait assessment tool are proposed. However these tools are not preferable among clinicians since they are expensive and interfere with ones´ normal gait pattern. This paper presents a simple, inexpensive, and effective system for extracting features from gait to distinguish abnormal gait from normal one. For classifying the normal and abnormal gait, first the silhouette and its bounding box are extracted in each frame. Then, the height of silhouette of different subjects is normalized. In the next step, the patterns obtained from different subjects are aligned temporally. These steps followed by tile extraction phase. Finally features are extracted from the gait frieze pattern. The method is evaluated by using for classifying 64 walking sequence (8 normal, 56 abnormal) to 8 different gait patterns. The results are shown that the method is able to determine the abnormal gait pattern with acceptable CCR.