Résumé

🚀 Objective & Vision

To develop a low-cost, anthropomorphic upper-limb prosthesis capable of real-time neural control — accessible to patients, research labs, and educational institutions.
More than a technical prototype, the project became a platform for interdisciplinary exploration, merging mechatronics, applied neuroscience, human-centered design, and assistive technology.

The goal was clear: turn intention into action, thought into motion, and technology into autonomy.

​

🛠️ Technical Architecture & Implementation

I led the complete system development — from mechanical modeling to functional neural integration.

• Learned how to prototype in CAD software (Maya), following human anatomical proportions and enabling realistic movement across the shoulder, elbow, wrist, and fingers.

• Fabricated the structure using 3D printing (PLA/PETG) for an optimal balance between lightness, durability, and affordability.

• Integrated servo motors for digital (finger) and analog (joint) control, ensuring proportional response and movement fluidity.

• Built the control system using an Arduino microcontroller, with supporting electronics including power drivers, voltage dividers, and signal filtering.

• Structured the system as a modular and expandable platform, ready to incorporate additional sensors, actuators, or alternate input methods.

​

📐 Mechanical Architecture & Assembly Insights

The robotic hand featured tendon-based tensioning mechanisms, actuated by servo motors that mimicked human muscular contraction and release.
Finger articulation was modeled after human joint movement, with torque distributed across custom-printed hinges and nylon cables.
During early field trials, tendon lines would snap under repetitive stress. This revealed load tolerances beyond our initial estimates and led to a redesign that doubled cable lifespan — one of several hidden thresholds uncovered through iteration.

​

🧠 Neuromuscular Interface & EMG Integration

Integrated surface electromyography (sEMG) technology using the Myo Armband to capture forearm muscle activity and translate it into real-time control signals.

• Developed algorithms to classify EMG signal patterns associated with distinct muscle contractions, enabling the triggering of specific motor functions in the robotic arm.

• Designed a custom calibration interface with visual feedback, allowing users to train, map, and fine-tune gestures or contraction thresholds to motor commands — adapting to each user's physiological profile.

• Implemented noise-reduction filters and signal-smoothing techniques to minimize false activations and ensure consistent responsiveness across different muscle groups and skin types.

​

💻 Signal Processing & Engineering Challenges

Built a complete EMG signal-processing pipeline from scratch, incorporating:

• Analog filtering (band-pass: 20–250Hz | notch: 60Hz) to eliminate environmental noise and powerline interference

• Signal amplification and rectification for peak detection

• Real-time digital smoothing (RMS + windowing) to stabilize gesture classification

• Efficient onboard computation using Arduino-based A/D conversion — optimized for latency without proprietary SDKs

​

🧬 Bio-Inspired Design Principles

The servo-motor layout mirrored the body’s own neuromuscular feedback loop — a kind of digital reflex arc.
Servo-motor mappings replicated muscle group synergy for naturalistic flexion and extension.
Control models were proportional, responding to the intensity — not just the presence — of muscle contractions. The aim was embodiment, not just functionality: a prosthetic that doesn't just move, but remembers how to move like a limb.

​

📐 Usability Testing & Interface Iteration

• Ran multiple bench and lab-based evaluations simulating user interaction

• Identified mechanical fatigue, signal drift, and joint tension inconsistencies

• Redesigned key mechanisms to reduce friction, optimize motion, and ensure durability

• Calibrated the EMG control loop with visual feedback to improve ease of use and response fidelity

​

🧭 Design Constraints & Engineering Trade-offs

• Chose Myo Armband for accessibility despite signal limitations compared to medical-grade EMG systems

• Prioritized modularity and openness over miniaturization or aesthetics

• Balanced servo torque and power draw to avoid overheating and maintain response speed

• Selected PLA and PETG for printability, despite wear resistance limitations — favoring accessibility and replicability

​

📊 Comparative Benchmarking

• Matched or exceeded gesture latency benchmarks (~200–250 ms) from published EMG control studies

• Achieved higher gesture clarity than low-cost commercial kits in the same price range

• Delivered complex multi-finger articulation for a total cost under USD ~$600, while many comparable systems exceeded $2,000

​

📁 Documentation & Knowledge Transfer

• Created internal documentation covering system assembly, wiring, and calibration workflow

• Produced schematic diagrams and CAD exports for reproducibility

• Served as a teaching tool in workshops and academic modules on biosignals, embedded systems, and prosthetics

​

🔍 Limitations & Engineering Constraints

Despite its success as a proof of concept, the project faced meaningful technical and mechanical challenges:

• Tendon-based tensioning introduced friction and required continual calibration to prevent loss of motion fidelity

• Gear-driven joints proved sensitive to repeated stress and required enhanced structural tolerance

• The low-level EMG-to-command layer, while innovative, increased system complexity and Bluetooth debugging overhead

These constraints directly informed redesigns and opened pathways for future iteration with new materials and embedded firmware.

​

🔮 Vision for Commercial Adaptation

With refinement in components and usability, this system presents a compelling base for:

• Low-cost EMG-controlled prosthetics in underserved communities

• Wearable neuroadaptive interfaces for smart home or IoT control

• Educational kits for teaching robotics, biosignals, and biomechanics

• A scalable platform for academic research in neuroprosthetics and assistive engineering

​

📰 Media Recognition & Public Impact

The project was featured in major Brazilian newspapers, including Estadão and O Tempo, highlighting its innovative use of muscle-controlled robotics to advance accessibility in assistive technology.
This media coverage positioned the project as both a scientific and social contribution — bridging research and public awareness.

​

🎓 Academic Contributions

The system was used as a teaching platform in courses and workshops on bioengineering, robotics, and signal processing.
It served as an accessible example of how biosignals, embedded systems, and design thinking can converge into functional, real-world solutions.

​

🧩 Research & Innovation Potential

• Proof-of-concept for real-time neuromechanical interfaces

• Foundation for low-cost prosthetics, wearable robotics, and gesture-controlled assistive tech

• Expandable into applications in rehabilitation, adaptive interfaces, and biomechanical education

• Architecture adaptable for future multi-modal sensing or machine learning integration

​

🧠 Reflections & Continuity

This project redefined how I think about autonomy — not as mechanical replication, but as restoring the ability to act, to decide, to move. It taught me that true innovation comes from listening: not just to code or components, but to the human signals underneath.
Today, I approach interface design not just as an engineer — but as a translator, turning intent into expression. This philosophy continues to guide my work in EMG Interfaces, Neurotechnology, and Human-Centered Robotics.