Parkinson’s disease (PD) is often characterised by distal tremor, affecting the patient’s hands and preventing them from performing daily tasks. There has been an increase in tremor suppression orthoses in the last decade, specifically containing active components that employ bio-signals for closed-loop control. However, most sensors tend to focus on the motion output for tremor detection, which is prone to inaccuracies and can hinder the responsiveness of the system. With the integration of machine learning in real-time tremor detection, this would enable the analysis of more complex signals, such as the muscle activation patterns of the subject’s forearm, which would give us indications on tremorgenic states based on the contractions rather than the resulting kinematics.This study therefore investigates the use of mechanomyography (MMG) as a sensing modality for muscle activity, and its capacity to detect tremor in a real-time setting for applications in tremor suppression devices. Wrist tremor is induced through transcutaneous electrical nerve stimulation, which enables the assessment of this approach as a proxy to PD tremor, and the ability of the MMG to work in conjunction with stimulation.Results show that the MMG-based algorithm was capable of effectively detecting when visible tremor was present, based purely on the muscle contraction patterns. This paves the way towards the use of MMG in closed-loop control systems of tremor suppression devices.