Toward practical mobile gait biometrics
Gait is the unique human locomotion due to individual specific biophysical and behavior habits. With ubiquitous mobile devices in people's daily life nowadays, accelerometers and gyroscopes provided in these devices directly capture the dynamic motion characteristics and thus have great potential for nonobtrusive gait biometrics. In fact, inertial sensors have been exploited to perform highly accurate gait analysis under controlled experimental settings. However, their performance in realistic scenarios is unsatisfactory due to variations in data measurements affected by physiological, environmental, and sensor-placement-related factors. Practical mobile gait biometric algorithms need to be robust to these variations to achieve high authentication performance in the field. It is the focus of this chapter to address some of these issues for in-the-wild mobile gait biometrics applications. First, we propose a novel gait representation called gait dynamics image (GDI) for accelerometer and gyroscope data sequences. GDIs are constructed to be both sensor-orientationinvariant and highly discriminative to enable high-performing gait biometrics for real-world applications. Second, we show how to further compute walking pacecompensated GDIs that are insensitive to variability in walking speed. Third, we adopt the i-vector paradigm, a state-of-the-art machine learning technique widely used for speaker recognition, to extract gait identities using the proposed invariant gait representation. Fourth, we demonstrate successful fusion of accelerometer and gyroscope modalities for improved authentication performance. Performance studies using both the naturalistic McGill University gait dataset and the large Osaka University gait dataset containing 744 subjects have shown dominant superiority of this novel gait biometrics approach compared to state-of-the-art. Additional performance evaluations on a realistic pace-varying mobile gait dataset containing 51 subjects confirm the merit of the proposed algorithm toward practical mobile gait authentication.
Toward practical mobile gait biometrics, Page 1 of 2
< Previous page Next page > /docserver/preview/fulltext/books/sc/pbse003e/PBSE003E_ch10-1.gif /docserver/preview/fulltext/books/sc/pbse003e/PBSE003E_ch10-2.gif