An EMG-driven virtual reality platform for physiology-based motor interaction: technical and functional validation

Harnessing experience-dependent neuroplasticity has become a central goal of modern neurorehabilitation. Immersive virtual reality (VR) has emerged as a promising tool, enabling task-oriented training in ecologically valid scenarios that are difficult to reproduce in conventional clinical settings. Despite these advances, interaction in most existing VR rehabilitation systems remains primarily mediated by external interfaces, such as motion tracking or handheld controllers, rather than by the user’s own physiological motor activity. Consequently, rehabilitation often fails to directly engage the physiological motor commands that naturally drive voluntary movement. We present and validate an integrated VR–electromyography (VR-EMG) platform that enables closed-loop, physiology-driven interaction between users and immersive virtual environments. The platform combines a standalone VR architecture with wearable EMG systems, online EMG decoding, and cloud-based performance monitoring within a unified framework. A standardized motor-command interface enables different EMG acquisition systems to control both anthropomorphic and non-anthropomorphic virtual effectors without modifying the decoding strategy or rehabilitation scenarios, providing a flexible framework for diverse applications. The platform was validated through complementary technical and functional evaluations performed on the two supported EMG acquisition systems. Technical validation assessed signal quality, gesture classification accuracy, communication latency, and system reliability. Functional validation was conducted in individuals with upper-limb amputation and able-bodied controls while performing object-interaction tasks across multiple immersive virtual scenarios. The platform demonstrated stable closed-loop operation, reliable gesture recognition, and the ability to quantify objective performance metrics, including task score, completion time, gesture classification errors, precision errors, correctly recognized gestures, and user-reported workload. Findings demonstrate the feasibility, technical reliability, and functional versatility of the proposed platform by providing a physiology-driven framework for immersive motor interaction that is compatible with different EMG acquisition systems and dissociates motor control from the morphology of the controlled virtual effector. Because the two systems were used for different components of the validation, we do not establish equivalence in their performance. By integrating online EMG decoding, VR, and quantitative performance monitoring, the platform offers a tool for future applications in rehabilitation, prosthetic training, and studies using functional magnetic resonance imaging and high-density electroencephalography to address whether training with the platform induces changes in functional brain organization.