Application of inertial sensors in the fall-risk assessment of elderly sanitation workers
Most sanitation workers in Chinese towns and villages are older with increasing fall risk.To enhance the safety level of sanitation workers,research on fall risk assessment using Inertial Measurement Unit(IMU)was conducted,and machine learning models were developed to assess the fall risk of older sanitation workers.Eighteen older sanitation workers participated in this experiment.Initially,the Dynamic Gait Index(DGI)assessment was used to determine the fall risk level of each subject,serving as the label of samples for machine learning.Five IMUs were used to collect the acceleration data of subjects'working-related motions,which were placed on each subject's sternum,pelvis,right upper leg,right knee,and right ankle.Simulating the common motions of sanitation workers:walking,squatting,and bending motion acceleration data were collected.The raw data was initially processed using a fourth-order low-pass Butterworth filter.Subsequently,Decision Tree,K-Nearest Neighbors,Naive Bayes,Support Vector Machine,and Optimizable Ensemble classifiers were employed to develop fall risk assessment models.The training dataset features included mean,range,variance,standard deviation,root mean square,skewness,kurtosis,and maximum Lyapunov index.Bayesian Optimization was used to optimize hyper-parameters of models.The Synthetic Minority Oversampling Technique(SMOTE)algorithm was applied to balance the sample data.The results demonstrated that the maximum accuracy reached 88.62％,the maximum F1 score was 90.00％,and the maximum AUC was 92.32％.In summary,the Support Vector Machine classifier,trained using right ankle acceleration data,exhibited the best overall performance with maximum values of accuracy and an F1 score of 90.00％.The results show that it is feasible to apply machine learning techniques to IMU-based fall risk assessment.Compared to previous studies,the optimized fall risk assessment model developed in this paper achieved an F1 score of over 95％,providing a better effect of assessment for older sanitation workers with high fall risk.The fall risk assessment model,based on cost-effective inertial sensing technology,is suitable for promoting elderly sanitation workers,thus improving their occupational safety.
NETL
NSTL
万方数据
