A Performance Comparison of Forcemyography and Electromyography in Muscle Force Estimation for Human–Machine Interfaces

The recording of muscle force can be used to control Human–Machine Interfaces (HMIs). A commonly used estimator of muscle force is the envelope of Electromyography (EMG) signals, which requires proper EMG processing. A potentially more convenient alternative is offered by Forcemyography (FMG), which records muscle mechanical activity and inherently provides a signal that is a good estimator of muscle force, without specific processing. This study aimed to compare the performances of EMG and FMG in muscle force estimation aimed at HMI control. Ten healthy subjects performed 15 gradual and 15 rapid isometric biceps brachii contractions at approximately 90° elbow flexion. Force measured at wrist by a load cell, and FMG and EMG signals were recorded simultaneously. Causal root-mean-square EMG envelopes (EMG-RMS) were computed offline using windows of 100, 250, 500, and 1000 ms. For each participant and contraction type, subject-specific sigmoidal regression models were fitted using normalized signals from the first 10 contractions and evaluated on the remaining five held-out contractions. Performance was assessed using the coefficient of determination (R2), root mean squared error (RMSE), and maximum absolute error (MaAE), with paired Wilcoxon tests and Bonferroni correction. Cross-correlation quantified temporal alignment with the reference force. FMG achieved higher mean R2 and lower mean RMSE and MaAE across all tested configurations. Statistical analyses on the errors revealed that FMG achieved performance comparable to that of EMG-RMS in gradual muscle contractions and significantly superior performance in rapid muscle contractions. These findings support FMG-based external force estimation with simpler pre-processing and provide preliminary evidence for future implementation and validation in muscle-controlled HMIs, assistive exoskeletons, and rehabilitation devices.