Hand-based human-machine interfaces (HMI) are complex tasks that involve repetitive and sustained movements and postures of the hand, that can cause muscle or tendon strain. Hence, for comfortable and safe use of these HMI, it is important to reduce the user's physical exertion. The evaluation of physical exertion can be performed either by the subjective evaluation of the relative perceived exertion or by objective physiological measurements. However, the relation between these two measures has not been sufficiently studied for localized low-exertion activities, as well as the combined effect of several active muscles on the level of perceived exertion in localized exertion activities.

This study investigated the relation between objective measures of exertion (handgrip force level and electromyography - EMG) and perceived exertion (Borg scale) during localized activities of forearm muscles. The effect of the combined activity of several muscles on the level of perceived exertion and the issue of gender differences in perceived exertion in these types of activities were studied as well.

Fifty four participants (27 females and 27 males) were recruited for three different experiments:​ two low-exertion hand gesture experiments (first and second experiments) and a moderate- to high-exertion handgrip force experiment (third experiment), during which Borg ratings of relative perceived exertion, handgrip forces and EMG signals from six forearm muscles were obtained.

Results revealed that females perceive low-exertion activities of the same relative intensity level (i.e., same muscle activity) as less effortful compared with males. For moderate- to high-exertion activities there was no significant gender difference in the perception of physical exertion. Additionally, males presented higher correlation between the objective and subjective measurements of exertion for moderate- to high-exertion activity, while there was no significant difference between the correlations of males and females for low-exertion activities.

Linear models for predicting the Borg ratings based on gender, and the combined activity of muscles, provided R-squared values ranging from 0.1 to 0.3 for lowexertion activities and from 0.42 to 0.51 for moderate- to high-exertion activity. The linear model based on gender and force level (for moderate- to high-exertion) provided R-squared value of 0.57. Moreover, when examining individual models (i.e., knowing the specific subject), the R-squared values were approximately 0.65 for low-exertion and up to 0.8 (up to 0.9 for the force model) for moderate- to high-exertion. Further, to test if the fitted linear models represent a general relation between EMG and Borg ratings, and therefore could be used to predict Borg ratings for different types of subject and physical activities, we used the models from the first experiment (i.e., models that were developed using the first experiment dataset) to predict the Borg ratings in the second and third experiments. Our analysis showed that by using the average values of muscles' activity (in other words, average EMG of the six recorded forearm muscles), we were able to achieve prediction abilities similar to those of the second and third experiment models (i.e., models that were developed based on the specific experimental dataset).

Finally, we tested our ability to predict the grip force level (i.e., applied force divided by the maximum force that the subject can generate) based on EMG signals and Borg ratings in moderate- to high-exertion activities. According to our analysis, when using each of the measurements separately, we were able to explain approximately 45% (R²≈0.45) of the variation in the estimated force level, while by combining both measures we were able to explain approximately 60% (R²≈0.6) of the variation. The results and the findings of this study can be used in monitoring perceived exertion (by using objective measurements only) during manual handling tasks or humanmachine interfaces. This could be of particular importance when it is not possible to obtain an oral perceived exertion rating, such as in cases where there is a task of high cognitive load where placing an additional load on the operator is not recommended (e.g., landing a plane, medical or machinery teleportation, etc.), or in instances where asking for the perceived exertion might cause a bias in judgment. Additionally, the models developed it this study could also enable predication of the absolute force level by using the maximum grip force for a given individual.