User recognition system and method thereof

A method of recognizing an input user from an input image, the method comprising:

• using a processor to perform
• a) capturing the input image of the input user using an image capture device and extracting a user feature vector from the input image captured by the image capture device;
• b) calculating a similarity between the extracted user feature vector and a user feature template of an enrolled user, the user feature template being user feature information that includes a plurality of clusters of the enrolled user, each of the clusters having a user feature vector as a member of the cluster, wherein the similarity is calculated by comparing a similarity between a centroid of each of the plurality of clusters of the user feature template and the user feature vector;
• and
• c) recognizing the enrolled user of the feature template as the input user when the similarity between the extracted user feature vector and the feature template exceeds a predetermined recognition threshold value,
• wherein each of the clusters represents a different facial condition of the same enrolled user,
• wherein when a given user is recognized in c), the method further comprises:
• d) capturing tracked input images of the recognized user after the recognized user has been determined to be an enrolled user by tracking the recognized user using the image capture device;
• e) extracting a feature vector from the captured image and re-verifying whether the captured image corresponds to an image of the recognized user;
• and
• f) when the input image is re-verified to be the same as the image of the recognized user in e), updating the user feature template of the recognized user with extracted feature vectors of the captured images
• wherein f) comprises:
• f-1) calculating a similarity between an extracted feature vector of the captured image and a cluster of the user feature template of the recognized user;
• f-2) searching for a winner cluster having a highest similarity, and calculating a similarity between the corresponding extracted feature vector and the winner cluster and the predetermined cluster classification threshold value;
• and
• f-3) including the corresponding extracted feature vector to the winner cluster as a member of the winner cluster when the similarity between the corresponding extracted feature vector and the winner cluster is greater than the predetermined cluster classification threshold value
• wherein after f-2), when the similarity between the corresponding extracted feature vector and the winner cluster is less than the predetermined cluster classification threshold value, the method further comprises generating a cluster and including the corresponding extracted feature vector to the generated cluster as a member of the generated cluster,
• wherein when a cluster is added and a total number of clusters exceeds a predetermined maximum number of cluster members, the method further comprises:
• selecting a cluster to be removed from among clusters of the user feature template based on at least more than one cluster having a highest mean similarity with other clusters and the least referred cluster in the user recognition process;
• and
• removing the selected cluster from the user feature template,
• wherein the cluster to be removed is selected by
• C
• d
• =
• arg
• ⁢
• ⁢
• max
• i
• =
• 0
• 1
• ⁢
• n
• ⁢
• (
• α
• ⁡
• (
• 1
• -
• C
• ri
• )
• +
• (
• 1
• -
• α
• )
• ⁢
• C
• si
• )
• wherein Cri is a contribution degree of the i-th cluster, Csi is a mean similarity between i-th cluster and other clusters, and α is a weight value.