Single-class classifier for identifying user identity, identification method, medium and computing equipment

Abstract

The invention relates to a single-class classifier for identifying a user identity, an identification method, a medium and computing equipment, and the method comprises the steps of carrying out the preprocessing of inertial data collected by a mobile phone sensor, and obtaining low-noise gyroscope and accelerometer data; acquiring posture change information of the user as a feature of user identity recognition; extracting statistical characteristics of the inertial data, and labeling the identity of a data collector; training a neural network as a feature extraction function, and performing feature clustering on the received statistical features by using a single-class classifier of MLP; training a multi-layer perceptron based on a contrast loss function, mapping the multi-layer perceptron to a hypersphere of a high-dimensional feature space, and increasing Euclidean distances between different pedestrian features; and finding out the center of the hypersphere, calculating the distance between the user characteristics and the center of the hypersphere, and realizing accurate authentication of the identity in the large-sample inertial data through a preset threshold value. According to the invention, the identity of the user can be identified with high reliability and high precision.

Description

technical field [0001] The invention relates to the technical field of biological feature identification and inertial navigation, in particular to a single-class classifier, identification method, medium and computing device for identifying user identity based on inertial navigation information. Background technique [0002] Micro-electromechanical inertial measurement units have the advantages of small size, light weight, and low power consumption, and have been widely used in wearable devices, smartphones, virtual reality, and health monitoring systems. Individual motion characteristics can be used as an important basis for biometric identification. Face recognition, fingerprint recognition and iris recognition are easily limited by light intensity, resolution, space and distance. The identification based on inertial data has the advantages of high concealment, strong autonomy, small amount of data, and fast operation speed. Existing recognition technologies are all base...

AI Technical Summary

Problems solved by technology

Existing identification technologies are all based on multi-classification methods, but in practical engineering applications, due to the large amount of user data and the diversity of features, traditional machine learning multi-classification methods cannot extract effective data feature sequences, resulting in a low identification accuracy rate. Low, it is not advisable to use classification for large-scale identification

Images