Put Your Muscle Into It: Introducing XEM2, a Novel Approach for Monitoring Exertion in Stationary Physical Exercises Leveraging Muscle Work

TL;DR

This work introduces XEM2, a camera-based system for measuring and visualizing muscle work at the level of individual muscle groups during stationary exercises. By integrating Azure Kinect-based motion capture, a Hill-type-inspired muscle model, and the Kinesis Unity plugin, it derives muscle-work values and presents them via a 3D avatar and limb-focused box plots. Technical validation (N=10, five exercises) demonstrates discriminability of muscle-work estimates, while a 36-participant user study shows XEM2 is perceived as informative and can complement traditional measures such as RPE and burned calories. The results suggest XEM2 offers granular, actionable feedback that supports balanced training and injury prevention, with future work aimed at improving accuracy, personalization, and mobile deployment.

Abstract

We present a novel system for camera-based measurement and visualization of muscle work based on the Hill-Type-Muscle-Model: the exercise exertion muscle-work monitor (\textit{XEM}$^{2}$). Our aim is to complement and, thus, address issues of established measurement techniques that offer imprecise data for non-uniform movements (burned calories) or provide limited information on strain across different body parts (self-perception scales). We validate the reliability of XEM's measurements through a technical evaluation of ten participants and five exercises. Further, we assess the acceptance, usefulness, benefits, and opportunities of \textit{XEM}$^{2}$ in an empirical user study. Our results show that \textit{XEM}$^{2}$ provides reliable values of muscle work and supports participants in understanding their workout while also providing reliable information about perceived exertion per muscle group. With this paper, we introduce a novel system capable of measuring and visualizing exertion for single muscle groups, which has the potential to improve exercise monitoring to prevent unbalanced workouts.

Figures (10)

• Figure 1: The human model represents the muscle groups we decided to visualize as they are primarily involved in the movement of upper and lower limbs (i.e., prime movers). The coloured areas on the human model are labelled with the muscle group name, while the colour indicates the affiliation to the four limbs.
• Figure 2: This Figure represents the implementation stack of the process from the movement to the XEM$^{2}$ visualization. On the left, there is a real-world human executing the movement. With the help of the Body Tracking SDK, the skeleton of the movement is transferred into a virtual model in Unity. In Unity, the Kinesis plugin transfers the movement into a muscle model where the forces of muscles are calculated and visualized as groups in the last highlighted 3D human model through XEM$^{2}$.
• Figure 3: The five exercises that were used to measure muscle work with XEM$^{2}$ for technical validity. Each exercise was executed 5 times in one measurement and five measurements were taken in total. Each measurement consisted of 5 executions of a movement, yielding a total of $5x5x5=125$ sampled motions.
• Figure 4: Plotted visualization of five exercises (arm circles, lunges, shoulder squeeze, squats with arms, squats without arms) measurements of XEM$^{2}$, showing values separately for each muscle group.
• Figure 5: The hardware setup for the study. On the left side is the 75 inches visualization screen ①, the Azure Kinect Camera ②, and a keyboard for text input ③. The right side depicts the participants' instrumentation: a participant wearing a Polar H10 heart rate sensor ④ and the Oculus Quest 2 headset, both for study part one ⑤.