Jun 2025 – Jun 2026
NeuroMotion
NeuroMotion is a real-time adaptive EEG brain-computer interface project I developed to detect motion intention from Motor Imagery (MI) and Movement-Related Cortical Potentials (MRCP). The system uses PyTorch-based CNN/RNN models, a TCP/UDP bridge between the Python AI backend and Unity 3D simulation, and an ErrP monitoring module to detect incorrect commands.
Overview
NeuroMotion is a real-time adaptive EEG brain-computer interface project I developed to detect motion intention from Motor Imagery (MI) and Movement-Related Cortical Potentials (MRCP). The system uses PyTorch-based CNN/RNN models, a TCP/UDP bridge between the Python AI backend and Unity 3D simulation, and an ErrP monitoring module to detect incorrect commands.
Problem
Motor-impaired users need control interfaces that can detect movement intention without conventional hand input, while EEG-based BCI systems must handle noisy signals, latency, and incorrect predictions.
Technical Approach
I built the AI pipeline around MI and MRCP signal interpretation with PyTorch CNN/RNN models, then connected the Python backend to a Unity 3D simulation through low-latency TCP/UDP sockets. The system also includes an ErrP module to monitor and correct mistaken commands.
Result
The project turns EEG motor-intention detection into an interactive assistive-control workflow, with the CV-documented communication bridge reducing end-to-end command latency to under 500ms.