Inertial sensor fusion for gait recognition with symmetric positive definite Gaussian kernels analysis

Abstract

Wearable sensor-based gait recognition has received much interest because it is unobtrusive and is user friendly. Many research has been carried out in this area but conventional gait recognition methods are not free from drawbacks. In this paper, accelerometer and gyroscope signals representing gait movements are encoded using covariance matrices. The covariance matrices provide a compact and descriptive representation for the accelerometer and gyroscope signals. Nonsingular covariance matrices are inherently Symmetric Positive Define (SPD) matrices. Interpreting such SPD matrices as points on the Riemannian manifold leads to increased performance. However, direct geodesic distance calculation for the matrix manifold may yield a suboptimal result. The proposed method solves this issue by embedding the manifold valued points to a higher dimensional Reproducing Kernel Hilbert Space (RKHS) via Positive Definite Gaussian Kernel functions. Extensive experiments have been conducted on three challenging benchmark datasets and a self-collected dataset. Experiment results testify the performance of the proposed RKHS embedding approach.

A fully connected deep learning approach to upper limb gesture recognition in a secure FES rehabilitation environment

“…32 In 2020, Permatasari et al's team optimized the gait recognition problem by using the accelerometer and gyroscope data to encode covariance matrices, and since nonstar covariance matrices are symmetric positive definite matrices, the SPD matrix was designed as a feature fusion of point pairs of data in the Riemannian-plane, which experimentally proved to be effective in overcoming the large data sets required in traditional gait recognition and the long computation time problem. 33 Zebin analyzed the effects of different parameters, features, and sensor locations on the overall recognition based on a model in which a wearable inertial sensor inputs a multichannel time-series signal and automatically outputs a classification of human body activities, and showed the importance of establishing a data set for different activities to classify activities of multiple genera. 34 Zhao et al used a multisensor to obtain lots of data to establish a self-supervised learning model for sleep recognition, increase data capacity through self-supervised pretraining, processing frequency domain information, use the rotational view t-stochastic neighbour embedding to represent multidimensional data features, and use the long short-term memory fusion condition random field, the test proved the effectiveness of the algorithm.…”

“…Segerra et al used sensor fusion data with gyroscopes and accelerometers to design an inertial data estimation orientation model based on Kalman filtering with Mahony filtering, which experimentally demonstrated the excellence of Kalman filtering, but upfront subjects had to be individually preprocessed to correct for bias 32 . In 2020, Permatasari et al's team optimized the gait recognition problem by using the accelerometer and gyroscope data to encode covariance matrices, and since nonstar covariance matrices are symmetric positive definite matrices, the SPD matrix was designed as a feature fusion of point pairs of data in the Riemannian‐plane, which experimentally proved to be effective in overcoming the large data sets required in traditional gait recognition and the long computation time problem 33 . Zebin analyzed the effects of different parameters, features, and sensor locations on the overall recognition based on a model in which a wearable inertial sensor inputs a multichannel time‐series signal and automatically outputs a classification of human body activities, and showed the importance of establishing a data set for different activities to classify activities of multiple genera 34 …”