Estimating Wrist Joint Angle Based on M-Mode Ultrasound Signals

Ultrasound sensing-based human-machine interface (HMI) is promising with the precise detection of the morphological changes of contracting muscles. M-mode ultrasound, a type of signal modulated through movement, has not yet been fully explored for HMI. This letter aimed to evaluate the capability of M-mode ultrasound for the continuous movement estimation of wrist joint. A commercial medical ultrasound system was employed to collect M-mode and B-mode ultrasound images simultaneously. Six able-bodied subjects were recruited for the wrist joint angle prediction. A feature extraction method of M-mode ultrasound was proposed. A strong linear relationship was found between the extracted features and the wrist joint angle. A linear regression model was adopted to compare the performance of two ultrasound modes under two different training strategies. The model was trained with training sets of different scales and complexity levels. For different training strategies, the M-mode ultrasound consistently kept the correlation coefficient (
r
) of over 0.9 and the normalized root mean square error (
NRMSE
) of below 0.15. Moreover, the M-mode ultrasound significantly outperformed B-mode ultrasound in the two training strategies. The results demonstrated the capability of M-mode ultrasound for wrist joint angle prediction, showing a great potential for the practical application in HMI.