Gait recognition using fuzzy ontologies and Kinect sensor data

Gait recognition involves the automatic classification of human people from sequences of data about their movement patterns. It is an interesting problem with several applications, such as security or medicine. Even low cost sensors can be used to capture pose sequences with enough quality to make a successful classification possible. In this paper, we describe the use of fuzzy ontologies to represent sequences of Microsoft Kinect gait data and some biometric features relevant for the gait recognition computed after them, enabling more reusable and interpretable datasets. We also propose a novel recognition algorithm based on fuzzy logic that outperforms state-of-the-art methods for straight line walks. We also face the problem of the identification of unknown individuals that are not present in the system knowledge base. © 2019 Elsevier Inc.