NeuroMesh: A Brain-Computer Interface Framework for Real-Time Motor Imagery Classification Using Edge-Optimized Deep Learning and Haptic Feedback Integration

Brain-Computer Interfaces (BCIs) represent one of the most transformative frontiers in human-computer interaction, yet widespread adoption remains constrained by computational latency, bulky hardware, and limited accessibility in resource-constrained environments. This paper presents NeuroMesh, a novel edge-computing Brain-Computer Interface framework designed and implemented as a B.Tech final-year project at Shri Sai College of Engineering and Technology (SSCET), Bhadravati. NeuroMesh enables real-time motor imagery classification from non-invasive electroencephalography (EEG) signals using an edge-optimized hybrid deep learning architecture combining lightweight Convolutional Neural Networks (CNNs) with Gated Recurrent Units (GRUs), deployed on the Raspberry Pi 4 and NVIDIA Jetson Nano platforms. The framework achieves end-to-end classification latency of 127 ms with 91.3% accuracy on the BCI Competition IV Dataset 2b, outperforming traditional cloud-dependent approaches that incur 800+ ms round-trip delays. A custom-designed haptic feedback subsystem translates classified motor imagery intentions into vibrotactile patterns delivered through wearable actuator arrays, enabling bidirectional human-machine communication without visual dependency. The system incorporates an adaptive calibration module using few-shot learning to personalize models to individual users within 5 minutes of initial setup, addressing inter-subject variability --- the principal challenge in EEG-based BCIs. All signal processing, feature extraction, inference, and feedback control execute locally on the edge device, ensuring complete data privacy and operability in offline environments. Rigorous evaluation across 12 participants demonstrates a System Usability Scale (SUS) score of 82.7, P300 spell-corrected communication throughput of 12.4 bits/min, and successful integration with assistive robotic actuators for upper-limb rehabilitation exercises.