A multi-algorithmic approach for gait recognition


Hind Al-Obaidi

Document Type

Conference Proceeding


Curran Associates Inc.


School of Science




Al-Obaidi, H., Li, F., Clarke, N., Ghita, B. & Keab, S. (2018). A multi-algorithmic approach for gait recognition. In A. Jøsang (Ed.), Proceedings of the 17th European Conference on Information Warfare and Security: ECCWS 2018 (pp. 20-28). Academic Conferences and Publishing International Limited.


Securing smartphones has increasingly become inevitable due to their massive popularity and significant storage and access to sensitive information. The gatekeeper of securing the device is authenticating the user. Amongst the many solutions proposed, gait recognition has been suggested to provide a reliable yet non-intrusive authentication approach - enabling both security and usability. Whilst several studies exploring mobile-based gait recognition have taken place, studies have largely been preliminary, with various methodological restrictions that have limited the number of participants, samples and type of features. Furthermore, prior studies have relied upon evaluating the approach on a limited number of activities - namely walking and running, and there is some concern over the capacity of the approach to correctly verify individuals when the nature of the signals across a wider range of activities is likely to be more variable. This paper has sought to overcome these weaknesses and provide a comprehensive evaluation, including an analysis of motion sensors (accelerometer and gyroscope), an investigation and analysis of features, understanding the variability of feature vectors during differing activities across a multi-day collection involving 60 participants. This is framed into two experiments involving five types of activities: normal, fast, with a bag, downstairs, and upstairs walking. The first experiment explores the classification performance of individual activities in order to understand whether a single classifier or multi-algorithmic approach would provide a better level of performance. The second experiment explored the features vector (comprising of a possible 304 unique features) to understand how its composition affects performance and for a comparison a more selective set of the minimal features are involved. Overall, results from the experimentation has shown an EER of 4.40/12.2%for a single classifier (using same/cross day methodologies). The multi-algorithmic approach achieved EERs of 0.70%/6.3%, 0.80%/12.68% and l.10%/6.46% for normal, fast and with a bag walk respectively (using the Same/ Cross Day methodology) using both accelerometer and gyroscope based features - showing a significant improvement over the single classifier approach and thus a more effective approach to managing the problem of feature vector variability.

Access Rights

subscription content