Calibrating open-source sEMG devices for accurate muscle signal analysis in sports and education

Abstract

Recent advancements in open-source technologies have considerably increased the appeal of low-cost devices for sports science professionals. However, achieving reliable measurements remains a major challenge. This paper introduces a straightforward yet robust calibration method for an open-source surface electromyography (sEMG) device, designed to balance affordable hardware with the rigorous accuracy requirements of sports science. In a controlled laboratory setting, artificial signals were generated to mimic the amplitude range (0.01–1 mV) and frequency range (up to 300 Hz) of real myoelectric activity. Both the sEMG device and an oscilloscope (serving as a reference gold standard) simultaneously recorded each signal. Regression analyses (ordinary least squares, OLS) coupled with paired t-tests, mean absolute error (MAE), and root mean square error (RMSE) were used to assess and quantify discrepancies across a wide array of amplitudes and frequencies. A strong linear correlation (R² = 0.994, p<0.01) was found between the sEMG readings and the oscilloscope values. Additionally, the model’s predictive accuracy was confirmed with a low mean absolute error (MAE) of 28.1 arbitrary units (a.u.) and a root mean square error (RMSE) of 37.67 a.u. These error metrics showed a high degree of agreement with the reference standard, enabling the derivation of a simple regression equation to convert raw sEMG data into consistent, reliable units. Compared to traditional methods focusing on internal amplifier or ADC calibration alone, this holistic approach calibrates the entire sEMG-ADC chain against a laboratory reference, ensuring that the final digitised signal accurately reflects true muscle activity. These findings demonstrate the true potential of applying open- source sEMG solutions to sports performance monitoring and education, while upholding stringent data quality standards.