Brain-Computer Interfaces (BCIs) enable users to interact with machines directly via neural activity, yet their real-world deployment is often hindered by bulky and powerhungry hardware. We present EdgeSSVEP, a fully embedded microcontroller-based Steady-State Visually Evoked Potential (SSVEP) BCI platform that performs real-time EEG acquisition, zero-phase filtering, and on-device classification within a lowpower 240 MHz MCU operating at only 222 mW. The system incorporates an 8-channel EEG front end, supports 5-second stimulus durations, and executes the entire SSVEP decoding pipeline locally, eliminating dependence on PC-based processing. EdgeSSVEP was evaluated using six stimulus frequencies (7, 8, 9, 11, 7.5, and 8.5 Hz) with 10 participants. The device achieved 99.17% classification accuracy and 27.33 bits/min Information Transfer Rate (ITR), while consuming substantially less power than conventional desktop-based systems. The system integrates motion sensing to support artifact detection and improve robustness and signal stability in practical environments. For development and debugging, the system also provides optional TCP data streaming to external clients. Overall, EdgeSSVEP offers a scalable, energy-efficient, and secure embedded BCI platform suitable for assistive communication and neurofeedback applications, with potential extensions to accelerometer-based artifact mitigation and broader real-world deployments.

翻译：脑机接口（BCI）允许用户通过神经活动直接与机器交互，但其在实际部署中常受限于笨重且高功耗的硬件。本文提出EdgeSSVEP，一个完全基于微控制器的嵌入式稳态视觉诱发电位（SSVEP）BCI平台。该平台在仅运行于222 mW的低功耗240 MHz MCU上，实现了实时脑电图（EEG）采集、零相位滤波及设备端分类。系统包含一个8通道EEG前端，支持5秒刺激时长，并在本地执行完整的SSVEP解码流程，消除了对基于PC处理的依赖。EdgeSSVEP使用六种刺激频率（7、8、9、11、7.5和8.5 Hz）对10名参与者进行了评估。该设备实现了99.17%的分类准确率和27.33比特/分钟的信息传输率（ITR），同时功耗显著低于传统的基于台式机的系统。系统集成了运动传感以支持伪迹检测，提升了实际环境中的鲁棒性和信号稳定性。为便于开发和调试，系统还提供了可选的TCP数据流传输功能至外部客户端。总体而言，EdgeSSVEP提供了一个可扩展、高能效且安全的嵌入式BCI平台，适用于辅助通信和神经反馈应用，并具备扩展至基于加速度计的伪迹抑制及更广泛实际部署的潜力。