1、2F-22025 Asia and South Pacific Design Automation Conference12025 Asia and South Pacific Design Automation ConferenceHeadset-Integrated Brain-Machine Interface for Mind Imagery and Control in VR/MR Applications Zhiwei Zhong,Yijie Wei,Lance Go,Yiqi Li,Jie GuNorthwestern University,Evanston,IL,USA2F-2
2、2025 Asia and South Pacific Design Automation Conference22025 Asia and South Pacific Design Automation ConferenceApplicationsVR/MR HeadsetsInteraction MethodsEEG waves Electroencephalogram(EEG)Brains electrical activity Brain-Machine Interface(BMI)Brain activity tracking Active control Reactive cont
3、rol Passive controlGamingNew Opportunities from BMI for VR/MRMovie/ShoppingOffice2F-22025 Asia and South Pacific Design Automation Conference32025 Asia and South Pacific Design Automation ConferenceMotivation and Challenges Cumbersome wearing High power for AI models Lack of low-latency computing su
4、pport Lack of real-time feedback controlVR headsetEEG electrode capDesktopImmersive virtual environment&brain activity combination2F-22025 Asia and South Pacific Design Automation Conference42025 Asia and South Pacific Design Automation ConferenceThis WorkT4T6O2T3T5PZCZGND The First SoC supporting m
5、ind imagery control for VR/MR Integration with VR headset and optimized channel placement ISA and AI architecture for general-purpose mind imagery tasks Teacher-Student CNN and sparsity enhancement for power savingLDO x 3SPIInput MemoryOutput vectorDigital CoreModel ControlWeight MemoryLNASAR ADCPGA
6、LPF16 Channel AFEResultsIIR Band Pass FilterTeacher CNNStudent CNNInstr.WeightReferenceGeneratorLDO x 3LDO x 3Scan IOClock Gen.Analog CorePE ArrayOutput Mem.OOKDataShifterSparsity ControlClk.GatingPseudo-resistorINNINPOUTNOUTPSwitch Caps2F-22025 Asia and South Pacific Design Automation Conference520