Biometric User Authentication

摘要: Methods are disclosed of authenticating a device user. A motion sensor 126, such as an accelerometer or gyroscope, captures data during user-induced motion. feature vector obtained from the data, possibly by obtaining ceptral coefficients 202, is inputted to – convolutional neural network 204 trained distinguish between vectors different users. output used determine whether matches expected authorized user pattern. In one embodiment (fig. 6), based on captured group training users not including The may be one-class support machine (SVM) binary classifier 206. another 7), earlier vectors. It reproduces outputs, and authentication includes determining there discrepancy input Another method comprises capturing both image capture device, comparing pattern, analysing three-dimensional facial structure present.