Monitoring of human exercise load is a hot area of wearable technology. The mainstream human motion calculation method predicts human motion through offline machine learning techniques. Personalized monitoring poses new challenges to the original learning model. Aiming at the detection of R peak value of motion load signal and the location of QRS complex, an improved R-peak detection algorithm based on adaptive threshold is proposed. The feature extraction method of motion load signal is introduced from multiple dimensions. The characteristics of time domain features and frequency domain features are analyzed. The time domain features and the eigenvalues of frequency domain features are defined and extracted. Experiments show that the proposed algorithm has higher detection accuracy. A multi-threshold-peak step algorithm is proposed and a motion state machine model is established. According to the periodic changes of the trunk movement and the arm swing acceleration in motion, the feature values are extracted, the step detection and the gait discrimination are performed, and the influence of external acceleration interference is excluded. The state machine adopts a nested structure, which is divided into two layers, a parent state and a child state, and calibrates the state transition condition. A verification method for the validity of feature parameters is proposed. The selected feature parameters are taken as an example to analyze and simulate the method. Simulation results show that the characteristic parameters in this paper are effective in the monitoring of the reasonableness of exercise load.

监视人体运动负荷是可穿戴技术的热门领域。主流的人体运动计算方法通过离线机器学习技术预测人体运动。个性化监控对原始学习模型提出了新的挑战。针对运动负载信号R峰值的检测和QRS波群的定位，提出了一种基于自适应阈值的改进的R峰值检测算法。从多个维度介绍了运动负荷信号的特征提取方法。分析了时域特征和频域特征的特征。定义并提取时域特征和频域特征的特征值。实验表明，该算法具有较高的检测精度。提出了多阈值峰步算法，建立了运动状态机模型。根据躯干运动和手臂摆动加速度在运动中的周期性变化，提取特征值，执行步距检测和步态识别，并排除外部加速度干扰的影响。状态机采用嵌套结构，将其分为父状态和子状态两层，并校准状态转换条件。提出了一种验证特征参数有效性的方法。以所选特征参数为例进行分析和仿真。仿真结果表明，本文的特征参数可以有效地监测运动负荷的合理性。