公开(公告)号：US20140341439A1

公开(公告)日：2014-11-20

申请号：US14280353

申请日：2014-05-16

Applicant: TATA CONSULTANCY SERVICES LIMITED

Inventor： Kingshuk CHAKRAVARTY ,  Aniruddha SINHA ,  Diptesh DAS ,  Rohan Banerjee ,  Amit Konar ,  Sudeepto Dutta

IPC: G06K9/00 ,  G06K9/62

CPC classification number: G06K9/00348 ,  G06K9/44 ,  G06K9/627

Abstract: Method(s) and system(s) for identification of an unknown person are disclosed. The method includes receiving skeleton data comprises data of multiple skeleton joints of the unknown person from skeleton recording devices. The method further includes extracting G gait feature vectors from the skeleton data. Further, the method includes classifying each gait feature vector into one of N classes based on a training dataset for N known persons and computing a classification score for each class. The method also includes clustering the training dataset into M clusters based on M predefined characteristic attributes of the known persons, tagging each gait feature vector with one of the M clusters based on a distance between a respective gait feature vector and cluster centers of M clusters, and determining a clustering score for each M cluster. The method further includes identifying the unknown person based on clustering scores and classification scores.

Abstract translation: 公开了用于识别未知人的方法和系统。 该方法包括从骨架记录装置接收包括未知人的多个骨骼关节的数据的骨架数据。 该方法还包括从骨架数据中提取G步态特征向量。 此外，该方法包括基于N个已知人员的训练数据集将每个步态特征向量分类为N类中的一个，并计算每个类的分类分数。 该方法还包括基于已知人员的M个预定义特征属性将训练数据集聚类成M个群集，基于各个步态特征向量与M个群集的簇中心之间的距离来标记每个步态特征向量与M个群集中的一个， 以及确定每个M簇的聚类分数。 该方法还包括基于聚类分数和分类分数识别未知人。

公开(公告)号：US09208376B2

公开(公告)日：2015-12-08

申请号：US14280353

申请日：2014-05-16

Applicant: TATA CONSULTANCY SERVICES LIMITED

Inventor： Kingshuk Chakravarty ,  Aniruddha Sinha ,  Diptesh Das ,  Rohan Banerjee ,  Amit Konar ,  Sudeepto Dutta

IPC: G06K9/00 ,  G06K9/44 ,  G06K9/62

CPC classification number: G06K9/00348 ,  G06K9/44 ,  G06K9/627

Abstract: Method(s) and system(s) for identification of an unknown person are disclosed. The method includes receiving skeleton data comprises data of multiple skeleton joints of the unknown person from skeleton recording devices. The method further includes extracting G gait feature vectors from the skeleton data. Further, the method includes classifying each gait feature vector into one of N classes based on a training dataset for N known persons and computing a classification score for each class. The method also includes clustering the training dataset into M clusters based on M predefined characteristic attributes of the known persons, tagging each gait feature vector with one of the M clusters based on a distance between a respective gait feature vector and cluster centers of M clusters, and determining a clustering score for each M cluster. The method further includes identifying the unknown person based on clustering scores and classification scores.

Abstract translation: 公开了用于识别未知人的方法和系统。 该方法包括从骨架记录装置接收包括未知人的多个骨骼关节的数据的骨架数据。 该方法还包括从骨架数据中提取G步态特征向量。 此外，该方法包括基于N个已知人员的训练数据集将每个步态特征向量分类为N类中的一个，并计算每个类的分类分数。 该方法还包括基于已知人员的M个预定义特征属性将训练数据集聚类成M个群集，基于各个步态特征向量与M个群集的簇中心之间的距离来标记每个步态特征向量与M个群集中的一个， 以及确定每个M簇的聚类分数。 该方法还包括基于聚类分数和分类分数识别未知人。