Youmin Joung, Sunkyoung Yu, Junbeom Kim, Soyoung Park, Kisoo Yu, Samsung Electronics

View Abstract Sign in for premium content


Wearable Global Navigation Satellite System (GNSS) device has several limitations to provide good positioning performance. System-On-Chip (SOC) in wearable devices is too small to concentrate all various components and circuits on, so the antenna and power have limitations. The attitude of GNSS antennas changes rapidly due to the additional dynamics, making it difficult to continuously track signals. It is necessary to design a wearable GNSS receiver design considering these characteristics. To overcome these limitations, various studies on wearable GNSS receivers have been proposed, ranging from antenna hardware design aspect to position performance optimization. As a wearable SOC GNSS provider, Samsung Electronics is also investigating various studies to improve the performance of GNSS solutions, and this study conducted a feasibility test of environmental context determination algorithm through machine learning. The proposed algorithm may be used to improve location performance as a result of optimization of navigation filter according to the environmental contexts. By analyzing difference in the wearable GNSS data between open-sky and urban environment, components of machine learning have been selected to determine the environment. Principal Component Analysis (PCA) method and Support Vector Machine (SVM) are used for the learning method. The post-processing test results show a success rate of 94%. Furthermore, the algorithm is close to 90% by applying algorithms to the real-time wearable OS. Using the predicted environment context, the navigation filter is designed to use each tuning value that optimizes navigation performance. The proposed algorithm provides a high success rate of 94% with a small computational load in Wearable GNSS device. Tracking loops and navigation algorithms can be optimized according to the predicted environment without changing hardware components. It is expected to help improve location performance even in low power mode.