Gait Recognition Using Wavelet Transform

Abstract

We describe a new method for Automatic GaitRecognition based around the use of wavelet descriptors that model the periodic deformation of human gait. Wavelet descriptors have been used successfully to model the boundary of static or moving, rigid-bodied objects, but many objects actually deform in some way as they move. Here, we use wavelet descriptors to model not only the object's boundary, but also the spatio-temporal deformations under which the object's boundary is subjected. We applied this new method to the Large Gait Database, compiled at the University of South Florida (NIST/USF) and found that the wavelet descriptors obtained for each person appear to be unique and can be used for recognition. Successful recognition rates of over 94% were obtained from the Large Gait Database.

Gait recognition by combining wavelets and geometrical features

“…Our proposed approach is found to be efficient and working with higher recognition accuracy when compared with the methods used by other researchers [2] where wavelet descriptors were used for feature extraction and the recognition rates were 85% with MLP and 91% with KNN for level 4 decomposition. …”

“…The wavelet descriptors are better than fourier, since they are able to catch small differences between patterns and the decomposition level four is considered to be the best compromise between compactness of representation and preservation of shape information [2]. It preserves both time and frequency information of a signal.…”

Mobile Match-on-Card Authentication Using Offline-Simplified Models with Gait and Face Biometrics

“…Those include: average, median, min, max, standard deviation (SD), and median absolute deviation (MAD) acceleration of individual axes and their magnitude [21], [33], root mean square (RMS) acceleration [33], mean-and zero-crossings [33], principal component coefficients of acceleration [34], [35], binned acceleration distribution [21], [24], [33], time between peaks [21], discrete cosine and fast Fourier transformation coefficients [36], [37], [38], [39], and Mel-and Bark-frequency cepstral coefficients [27], [33]. Further, wavelet transformations have been used with non-cycle-based acceleration gait data [27], [40] and floor sensor based gait data [41], as well as on acceleration based gait style recognition [42], which in contrast to gait identification or authentication does not distinguish individuals but gait styles. On those features, again a number of non-DTW based models have been applied, including cross-correlation based [43] or tree based models [21], artificial neural networks (ANN) [21], [44], support vector machines [33], [35], analysis of variance (ANOVA) [36], Gaussian mixture models (GMM) [38], and hidden Markov models (HMM) [33].…”