May the Force Be with You: Dexterous Finger Force-Aware VR Interface

Advances in virtual reality (VR) have reduced experience differentials for users. However, gaps between reality and virtuality persist in tasks that require coupling users' multimodal physical skills with virtual environments in delicate ways. User embodiment in VR easily breaks when physicality feels inauthentic, especially when users invoke their innate predilection to touch and manipulate things that they encounter. In this research, we examine the potential of forceaware VR interfaces for enabling natural connections to user physicality and evaluate them in high-finesse cases of touch. Combining surface electromyography (SEMG) with visual tracking, we develop an end-to-end learning-based system, ForceSense, to decode users' dexterous finger forces from their forearm sEMG signals for direct usage in standard VR pipelines. This approach eliminates the need for hand-held tactile equipment, thereby promoting natural embodiment. A series of user studies on manipulation tasks in VR validate that ForceSense is more accurate, robust, and intuitive than alternative solutions. Two proofs-of-concept VR applications, calligraphy and piano playing, demonstrate that the good synergy between visual, auditory, and tactile modalities, as ForceSense affords, has the potential of enhancing users' task learning performance in VR. Our source code and trained models are released at https://github.com/NYU-ICL/vr-force-aware-multimodal-interface.