EMG-Controlled Hidden Blade – V1

This is a functional prototype of a wearable, spring-loaded, servo-triggered arm blade inspired by the Assassin's Creed hidden blade — controlled using EMG (electromyography) signals from forearm muscle flexion.

🧠 Concept

• Blade is mounted on a sled inside a forearm-mounted housing
• A compression spring pushes the sled/blade forward
• A servo-controlled latch holds the sled in place until triggered
• EMG signal (via MyoWare 2.0 sensor) detects muscle flex
• When threshold is passed, servo releases latch and blade extends

⚙️ Current Version

• ✅ Spring-based extension only
• ❌ No automatic retraction (planned for V2)
• ✅ EMG-to-servo prototype logic
• ✅ Breadboarded servo response with simulated input
• ✅ BiLSTM-CNN gesture classification system
• ✅ Real-time web dashboard for EMG visualization
• 🛠️ In-progress Onshape design for housing and blade sled

📐 Hardware Plan

Component	Status
MyoWare 2.0 EMG Sensor	Ordered
Servo Motor	Ready
Spring System	Designing
Blade Sled & Housing	CAD WIP
Microcontroller	Arduino Uno

💻 Software Architecture

Arduino Firmware (Release.ino)

Simple EMG threshold detection for servo trigger control.

Python Gesture Recognition (emg_gesture/)

Advanced BiLSTM-CNN neural network for multi-gesture classification:

emg_gesture/
├── data/           # EMG collection & preprocessing
├── model/          # BiLSTM-CNN architecture & training
├── inference/      # Real-time classification & actuator control
├── frontend/       # Web dashboard with live visualization
├── train.py        # Train the model
├── run.py          # Run real-time inference
└── dashboard.py    # Launch web dashboard

Features:

• 8-channel EMG signal processing (bandpass filter, 60Hz notch)
• Time-domain feature extraction (MAV, RMS, WL, ZC, SSC)
• Hybrid CNN (spatial) + BiLSTM (temporal) architecture
• Real-time gesture classification: closed_hand, open_hand, pointing
• Serial actuator control with gesture-to-command mapping
• Cyberpunk-themed web dashboard

Quick Start

cd emg_gesture
pip install -r requirements.txt

# Train model (generates mock data for testing)
python train.py --generate-mock --epochs 20

# Launch dashboard
python dashboard.py

# Run CLI inference
python run.py --mock

🛠️ Build Goals

• CAD housing + sled
• Design spring + latch system
• Wire EMG sensor to controller
• Test trigger + safety threshold logic
• Build gesture classification model
• Create real-time dashboard
• Final assembly and arm mounting
• Integrate classifier with Arduino

📸 Progress Log

Week 1 – Pre-Build Prep

• Created repo, sketched sled/latch system
• Simulated EMG input using potentiometer
• Breadboarded servo control logic
• Created Onshape base housing model

Week 2 – Software Development

• Built complete EMG gesture recognition pipeline
• Implemented BiLSTM-CNN hybrid model
• Created real-time web dashboard
• Added mock mode for hardware-free testing

📦 Parts List

• MyoWare 2.0 Muscle Sensor Basic Kit
• 3M Red Dot Electrodes
• SG90 or MG90S Servo
• Springs (compression, medium stiffness)
• Arduino Uno

---

🤖 Future Plans

• Add automatic retraction (winch or linear actuator)
• Integrate battery + wireless control
• Create 3D printed wearable housing
• Explore gesture-based EMG classification ✅ Done!
• Train on real EMG data from MyoWare sensor

---

⚠️ Disclaimer

This project is for educational and prototyping purposes only. It uses a spring-loaded mechanism and should be handled with extreme care. Do not use it in public or without proper safety measures.