Recurrent Neural Network for Inertial Gait User Recognition in Smartphones

Abstract

In this article, a gait recognition algorithm is presented based on the information obtained from inertial sensors embedded in a smartphone, in particular, the accelerometers and gyroscopes typically embedded on them. The algorithm processes the signal by extracting gait cycles, which are then fed into a Recurrent Neural Network (RNN) to generate feature vectors. To optimize the accuracy of this algorithm, we apply a random grid hyperparameter selection process followed by a hand-tuning method to reach the final hyperparameter configuration. The different configurations are tested on a public database with 744 users and compared with other algorithms that were previously tested on the same database. After reaching the best-performing configuration for our algorithm, we obtain an equal error rate (EER) of 11.48% when training with only 20% of the users. Even better, when using 70% of the users for training, that value drops to 7.55%. The system manages to improve on state-of-the-art methods, but we believe the algorithm could reach a significantly better performance if it was trained with more visits per user. With a large enough database with several visits per user, the algorithm could improve substantially.

Vein Biometric Recognition on a Smartphone

“…Smartphones are essential in daily life and are the systems with the most biometric modalities embedded simultaneously on the same device: fingerprint (e.g., Vivo© X20 [6]), facial (e.g., Huawei© Mate 20 Pro [7]), iris (e.g., Samsung© Galaxy S8 [8]) and voice (e.g., Google Assistant [9]). Likewise, in the research world, other modalities for smartphones are studied: gait [10], keystroke [11], or handwriting [12].…”

Data Augmentation for Inertial Sensor-Based Gait Deep Neural Network

“…The first gait recognition study using inertial sensor was conducted by Ailisto et al [5]. Offering many attractive advantages (e.g., small size, mobility, low cost, implicit operation), this approach quickly received significant research effort after that, and achieved promising results [6]- [8], [14]- [16], [34]. In the beginning stage, conventional machine learning and pattern recognition techniques were adopted to discover from raw data the meaningful feature/template for user identification/verification.…”

“…Recently, deep learning has gained an extraordinary development and dramatically improved the state-of-the-art researches in many pattern recognition and machine learning tasks such as speech recognition, visual object recognition/detection [35]. Following that trend, many studies have adopted deep leaning techniques for the task of inertial sensor-based gait recognition and achieved new state-of-theart results [9]- [12], [14]- [16], [36].…”

“…The CNN then used AE-GDI image as the input for predicting the owner's identity. On the newest research [16], recurrent neural network (RNN) was firstly adopted on this task, and a promising result was reported.…”

“…Although deep learning techniques have been applied and established new state-of-the-art researches in this field recently [9]- [12], [14]- [16], the scarcity of training data still endures as an obstacle that impedes to build a robust deep gait model [7]. Specifically, training a deep network usually requires a large and diverged dataset which covers a substantial number of data instances possibly occurring in practice [17], [18].…”