Method for identifying a subject using gait analysis
AssigneeKing Fahd University of Petroleum and Minerals
InventorsAmer Ghazi Abdullah Binsaadoon, El-Sayed M. El-Alfy
Life of this patent
Estimating the term…
Abstract
Described is a novel method for feature extraction for automatic gait recognition. This method uses Multi-kernel Fuzzy-based Local Gabor Binary Pattern. From a captured gait video sequence, the gait period is determined then a gait energy image is constructed to represent the spatial-temporal variations during one motion cycle of the gait sequence. Then, each gait sequence is represented with a feature vector. The computation of this vector is conducted by first applying the 2D Gabor filter bank then encoding the variations in the Gabor magnitude using a multi-kernel fuzzy local binary pattern operator. Finally, gait classification is performed using a support vector machine.
Claims5
A method of subject recognition by a classification model trained based on gait, performed by executing a program by one or more processors in a computer system, the method comprising:
- obtaining a video sequence of a subject walking;
- determining a set of feature vectors from the video sequence of the subject walking;
- training the classification model with the set of feature vectors;
- and
- classifying a new video sequence in order to recognize the subject walking,
- wherein the determining the set of feature vectors comprises:
- inputting, into the computer system, the video sequence of the subject walking;
- determining, by the one or more processors, an average image of images of the video sequence for a cycle;
- filtering, by the one or more processors, the averaged image to obtain N filtered responses, N is a natural number greater than one;
- determining, by the one or more processors, the magnitudes of the filtered responses for a center pixel pc and a selected plurality of neighbor pixels having a radius r;
- for each filtered response, determining for each of K different radiuses from the center pixel pc, by the one or more processors, a plurality of fuzzy based local patterns based on the magnitudes of the filtered responses, including applying a kernel fuzzy local binary pattern operator that sums magnitude differences of the filtered response for selected pixels at a respective same radius r from the center pixel pc, to generate the plurality of fuzzy-based local patterns for each of the respective K different radiuses, K is a natural number greater than one;
- for each filtered response, summing the fuzzy-based local patterns for each of the respective kernel fuzzy local binary pattern operators to form a histogram for a respective filtered response;
- forming, by the one or more processors, feature vectors based on the histograms of the N filtered responses;
- and
- identifying the subject walking based on the filtering and the determining.
Dependencies are inferred from the claim text.