Master of Engineering · 2025
A wearable sensing network that translated hand gestures into visible and spoken letters.
System developed
Five flex sensors and a six-axis IMU captured finger posture and hand movement. An ESP32 conditioned and streamed the data to a Python recognition pipeline with networked display and text-to-speech output.
Project storyline
Five flex sensors measured individual finger bend while an MPU6050 captured orientation and movement. An ESP32 at the wrist sampled both sensing modes and prepared the signals for live transmission.
The ESP32, inertial sensor, wiring, and finger sensors were integrated directly into the glove. Placement and strain relief mattered because the electrical system needed to remain stable through repeated hand movement.
Exponential moving averages, a small dead band, fixed ADC settings, and IMU filtering reduced jitter. Per-finger rest and full-bend calibration adapted the glove to sensor and fit variation.
Five normalized bend values and three accelerometer axes formed an eight-feature vector. Python and scikit-learn were used to compare k-nearest neighbors, Random Forest, support vector machine, and multilayer perceptron models. kNN was selected for the live pipeline.
The Python interface displayed each recognized letter and could speak it aloud. Any device on the same Wi-Fi network could open the shared live interface. The complete alphabet, including dynamic J and Z gestures, was included in the demonstration.
Aaron Emmanuel · George Mikhaiel
MIE1050H · Design of Intelligent Sensor Networks