| Skeletal muscle is an important biological tissue that maintains the posture of the human body and realizes human movement.Therefore,the estimation of muscle strength(MS)is of great significance in the fields of sports training,rehabilitation engineering and medical auxiliary equipment.The mechanomyography(MMG)is a vibration signal generated when the muscle contracts,and can reflect the mechanical characteristics of the muscle.Therefore,MMG was used in this subject to study the estimation of triceps MS.The purpose of this subject is to analyze the correlation between MMG and MS,and establish a widely applicable triceps MS estimation model.In this study,ADXL354 triaxial accelerometer,ZNLBS-VI-30KG tension sensor,and NI9205 acquisition card were selected to realize the synchronous acquisition of MMG and MS signal,and an experimental data acquisition interface was designed.Subsequently,the subjects’ constant force contraction experimental data at different MS levels were collected,and the relationship between MMG features and MS was fitted and compared using linear,quadratic polynomial,and exponential functions.Finally,the MMG of the subjects under different contraction patterns was collected through variable contraction experiments.The MS estimation effects of three models including BP neural network,extreme gradient boosting(XGBoost),and quadratic polynomial were compared and analyzed to select the final MS estimation model,and verified the applicability of the MS estimation model in different contraction patterns.The results show that there is a non-linear relationship between the features of MMG and MS(p<0.05);Although the MS estimation effect of the quadratic polynomial model is slightly worse than the XGBoost model,it has great advantages in the field of engineering applications,so the quadratic polynomial was selected as the final MS estimation model,and the goodness of fit between estimated MS and actual MS reaches 0.9178±0.0354,and the MS estimation model has universal applicability to the MS estimation under different contraction patterns(p<0.05). |
