Project
| # | Title | Team Members | TA | Documents | Sponsor |
|---|---|---|---|---|---|
| 10 | NeuroBand |
Arrhan Bhatia Vansh Vardhan Rana Vishal Moorjani |
|||
| # Problem As LLM-based voice assistants move onto AR glasses, interacting by voice is often impractical in public (noise, privacy, social norms). Existing AR inputs like gaze/head pose can be fatiguing and imprecise for pointer-style tasks, and camera-based hand-tracking ties you to specific ecosystems and lighting conditions. We need a device-agnostic, silent, low-latency input method that lets users control AR (and conventional devices) comfortably without relying on voice. # Solution Overview We propose a two-band wrist/forearm mouse that connects as a standard Bluetooth HID mouse and operates in virtual trackpad mode: * A wrist band (Pointing Unit) uses an IMU to estimate pitch/roll relative to a neutral pose and maps that orientation to a bounded 2D plane (absolute cursor control). A clutch gesture freezes/unfreezes the cursor so the user can re-center their wrist naturally. * A forearm band (Gesture Unit) uses surface EMG electrodes over the forearm muscle belly to detect pinch/squeeze gestures for clicks, drag, right-click, and scroll. * The wrist band is the host-facing device (Bluetooth HID). The forearm band communicates locally to the wrist band (tether or short-range wireless) for low added latency. * Initial design focuses on pitch/roll; yaw is not required for trackpad mode. # Solution Subsystems ## 1 — Wrist Band (Pointing Unit) * Wrist-mounted inertial sensing to estimate stable pitch/roll relative to a neutral pose. * Lightweight fusion/filtering for smooth, low-noise orientation signals suitable for absolute cursor mapping. * Local state for clutch (engage/hold/release) and pointer acceleration/limits as needed. ## 2 — Forearm Band (Gesture Unit) * Noninvasive EMG sensing over forearm muscle groups associated with finger pinches. * Basic signal conditioning and thresholding to convert muscle activity into discrete actions (left click, right click, drag, scroll). * Brief per-user calibration to set comfortable sensitivity and reduce false triggers. ## 3 — Inter-Band Link & Firmware * Local link from the forearm band (gesture events) to the wrist band (pointing and HID reports). * Embedded firmware to read sensors, perform fusion/gesture detection, manage clutch, and assemble standard Bluetooth HID mouse reports to the host. * Emphasis on responsiveness (low end-to-end latency) and smoothness (consistent cursor motion). ## 4 — Power, Safety, and Enclosure * Rechargeable batteries and simple power management sized for day-long use. * Electrical isolation/protection around electrodes for user safety and comfort. * Compact, comfortable bands with skin-safe materials; straightforward donning/doffing and repeatable placement. # Criterion for Success * Pairs as a standard BLE mouse and controls the on-screen cursor in virtual trackpad mode. * Supports left click, right click, drag, and scroll via gestures, with a working clutch to hold/release cursor position. * End-to-end interaction latency low enough to feel immediate (target: sub-~60 ms typical, Apple's magic mouse 2 has a latency of ~60 ms before motion is reflected on screen). * Pointer selection performance on standard pointing tasks comparable to a typical BLE mouse after brief calibration. * Minimal cursor drift when the wrist is held still with clutch engaged. * High true-positive rate (>= 90%) and low false-positive rate for click gestures during normal wrist motion. * 4 hours of battery life on a single charge. * Stable wireless operation in typical indoor environments at common usage distances (up to 2 meters). |
|||||