The Turning Point
During my six-month internship at the Translational Neural Engineering (TNE) laboratory of EPFL, I completed my master's thesis. This work laid the foundation for everything that would follow in my career.
Part 1: Modular Architecture for Biomedical Sensors
The Problem
Existing devices like the OT Biolab 400-channel EMG sensor weren't compatible with other software. Researchers needed a solution that could integrate and manage data from multiple sensors.
The Solution
I designed and implemented a modular architecture that could:
-
Process and record real-time data from various biomedical sensors
-
Work on both Windows and Linux
-
Use Lab Streaming Layer for network communication
-
Allow monitoring and analysis from multiple sources
Part 2: Correlating EMG Electrodes with Movement Decoding
Research Objectives
-
Evaluate how the number and position of electrodes affects decoding quality
-
Assess inter-session and inter-subject relationships of predictive models
-
Apply LDA for classification and LSTM neural networks for regression on finger movements
Key Findings
Through experiments with three subjects over several weeks:
-
16-26 electrodes significantly improved decoding quality compared to 6 electrodes
-
Diminishing returns with higher numbers of electrodes
-
Electrode position matters as much as quantity
The Significance
This thesis answered a fundamental question: How do we decode human intention from muscle signals?
The answer would become the core of my work on prosthetic hands. If we can reliably decode what a person intends to do from their muscle activity, we can build prosthetics that respond to thought rather than requiring conscious effort.
Career Trajectory
This project opened doors to:
-
Research positions in prosthetics
-
Deep understanding of EMG signal processing
-
Machine learning applied to biomedical problems
-
The CleverHand project that would define my career