EMG Waveforms
Muscle Activation
Real-Time Prediction
Feature Heatmap
EMG Waveform Monitor
Full-resolution EMG signal from all 8 channels. Signals are filtered through a 50 Hz notch filter before feature extraction.
Gesture Training
Click a gesture card to record a 3-second sample. Collect at least 10 samples per gesture before training.
Machine Learning
Confusion Matrix
Calibration Wizard
Channel Sensitivity
Servo Control
Manual Override
Parameters
Active Gesture
Finger Control
Independently control each finger servo from 0° (closed) to 180° (open).
3D Hand & IMU Telemetry
Model Training
Train a Random Forest or Neural Network classifier on collected EMG gesture samples.
Confusion Matrix
Classification accuracy per gesture after training.
Session Recording
Saved Sessions
Deploy to Teensy Standalone Mode
Run Without a Laptop (Standalone Mode)
Bake your trained classification model and calibration parameters directly into native Teensy 4.1 firmware. Once compiled and flashed, the ARMx physical prosthesis executes gesture recognition calculations entirely on-chip, requiring no active USB telemetry streaming, no Python host processing, and no external computer connection.
Technical Specifications
- Decision trees compiled to native, high-performance C++ (executes at <20 ms per inference loop).
- Active calibration limits and sensor parameters are embedded dynamically as static float arrays.
- Firmware compilation overwrites previous deployments securely in a single flash operation.
- Normal laptop streaming mode is completely unaffected—both firmware behaviors coexist natively.
Requires a fully trained gesture model. Sensor calibration is recommended prior to compilation.
Flash Instructions
- Click the Deploy to Teensy action button to compile the C++ source files.
- Open the generated
firmware/standalone/standalone.inosketch in Arduino IDE. - Select Teensy 4.1 from the Arduino Boards menu.
- Click Upload in the IDE. The arm will restart immediately in standalone mode.
- To revert back to active laptop streaming telemetry mode, re-flash
emg_reader.ino.
System Firmware
Upload these firmware sketches to your Teensy 4.1 microcontroller using the Arduino IDE.
🦾 Main EMG Reader
Required for real-time telemetry streaming, muscle activation plotting, and laptop control mode.
Download emg_reader.ino🎛️ Servo Limit Tuner
Utility script to safety-test physical range of motion, prevent motor burnouts, and calibrate limits.
Download servo_tuner.inoManuals & User Guide
Detailed guides for hardware pinning, muscle sensor placement, proportional control parameters, and API wiring.
📄 Setup Guide & Manual
Comprehensive guide including Teensy 4.1 hardware wiring schematics, calibration steps, and troubleshooting FAQs.
Download Setup Guide🔌 API Reference Specification
Endpoints schema reference for serial handlers, REST parameters, and real-time WebSockets event payloads.
Download API SheetARMx Panel Reference & Operation Manual
Welcome to the revised ARMx operation manual. Below is a detailed functional breakdown of each core panel within the diagnostic interface, updated to reflect our recent layout and visual upgrades.
🖥️ Main Dashboard & Signal Monitoring
- 📈 EMG Waveforms & Muscle Activation Streams raw multi-channel EMG data from the Teensy 4.1. The Muscle Activation panel plots filtered muscular contraction percentages in real time, featuring interactive threshold bars and raw activity logs.
- 🔮 Real-Time Prediction HUD (Uplifted) Displays active machine learning classifications using a custom scifi-inspired scanning beam overlay and a pulsating gesture badge. Includes a compact 2-column grid mapping for all 9 classifier output channels, ensuring real-time confidence scores are highly legible and visually balanced without vertical card elongation.
- 🧮 Feature Heatmap & Confusion Matrix Computes and renders real-time signal feature metrics (RMS, MAV, VAR, ZCR, WL, SSC) across all 8 muscle channels in a neat, interactive matrix. Generates real-time classification performance charts once a model is loaded.
🎛️ Calibration, Training & Manual Overrides
- 🎯 Proportional Calibration Panel Automates dynamic baseline adjustments. Guides the operator through baseline resting, peak contraction, and active sensor scaling. Features individual channel gain adjustments.
- 🦾 Hardware Control & 3D hand Telemetry (Relocated) Relocated to the Control Page to create a comprehensive manual testing environment! Places manual servo sliders, gesture quick-override buttons, and the **Polished 3D Peach-Clay Hand Visualizer** side-by-side. The 3D clay model features rounded side-contours, smooth muscle eminences, a bulge-free flat wrist transition, and a reframed camera view to guarantee extended fingers never clip the circular frame.
- 🚀 Training & Hardware Deployment Enables EMG dataset recording (saving baseline feature matrices for REST, OPEN, FIST, and other presets). Triggers backend SVM/ANN training and compiles standalone compiled firmware code for instant, host-free Teensy execution.
System Settings
System configuration and preferences.
About ARMx
ARMx is an open-source EMG-controlled prosthetic arm system built around the Teensy 4.1 microcontroller. It provides real-time gesture classification, calibration, training, and standalone deployment.
Made with pride by Digvijay Singh Murmu
Security & Access
Protect your dashboard settings, training sets, and physical servo commands with an encrypted access lock.
👥 Registered Operator Accounts
View, edit, reset passwords, or delete registered operator accounts.
| Operator | Role | Bio | Datasets / Models | Created | Actions |
|---|---|---|---|---|---|
| Loading operators... | |||||
📂 Operator Data & ML Inspector
Inspect, download, or delete individual training sessions and machine learning models of active operators.
📣 Publish Notification Update
Broadcast a system update or feature log directly to the Operator timeline.
📋 Current Update Logs
Manage the updates list served to operator timeline.