GNSS and Physiological Data for Running Monitoring

Motion analysis forms a very important research topic with a general mathematical background and applications in different areas including engineering, robotics, and neurology. This paper presents the use of the global navigation satellite system (GNSS) for detection and recording of the moving body position and the simultaneous acquisition of signals from further sensors. The application is related to monitoring of physical activity and the use of wearable sensors of the heart rate and acceleration during different motion patterns. The proposed computational methodology  includes digital methods of data preprocessing, time synchronization, and data resampling to enable their correlation, feature extraction both in time and frequency domains, and classification. The datasets includes signals acquired during ten experimental runs in the selected area. Detection the motion patterns includes segmentation of signals using analysis of GNSS data, pattern evaluation, and classification of motion signals in different terrain conditions. Classification methodology provides the comparison of neural networks, support vector machine, Bayesian, and $k$-nearest neighbour methods with the highest accuracy achieved by combined features and the two layer neural network. The proposed methodology and graphical user interface suggest the efficiency of the multi-channel and multi-dimensional signal processing with applications in rehabilitation, fitness movement monitoring, neurology, engineering, and robotic systems as well.运动分析是一项兼具通用数学理论基础与多领域应用价值的重要研究课题，其应用场景涵盖工程、机器人学与神经病学等方向。本文提出采用全球导航卫星系统（Global Navigation Satellite System, GNSS）检测并记录运动物体的位置，同时同步采集其他附加传感器的信号。该应用聚焦于身体活动监测，以及在不同运动模式下使用采集心率与加速度的可穿戴传感器。本文提出的计算方法包含数据预处理、时间同步与数据重采样等数字化手段，以实现多源数据关联、时域与频域特征提取，以及分类任务。本数据集包含在选定区域内开展的十次实验采集过程中获取的信号。运动模式检测环节包括基于GNSS数据分析的信号分割、模式评估，以及不同地形条件下的运动信号分类。分类方法部分对比了神经网络、支持向量机、贝叶斯方法与k近邻（k-nearest neighbour）算法的性能，其中结合多特征与双层神经网络的方案实现了最高分类精度。本文提出的计算方法与图形用户界面（Graphical User Interface, GUI）验证了多通道多维信号处理的有效性，其应用场景涵盖康复医学、健身运动监测、神经病学、工程学与机器人系统等领域。