Abstract
A data glove provides a wearable human-computer interface by measuring finger bend and hand orientation. This design combines five flex sensors, an MPU6050 inertial sensor, an HC-05 serial Bluetooth module, a TFT display, and an STM32F103 controller. Firmware written in C runs tasks under µC/OS-II and sends processed gesture data to a robotic arm or virtual hand.
Hardware design
Each flex sensor forms part of a voltage-divider circuit connected to an ADC channel. Its resistance changes with bending, producing a finger-specific measurement. Because placement and users differ, raw values require per-finger calibration for relaxed and bent poses.
The MPU6050 combines a three-axis accelerometer and three-axis gyroscope. It communicates over I²C and can provide raw motion samples or processed orientation data through its digital motion processor. The HC-05 transports compact frames over a serial link, while the TFT presents connection, calibration, and gesture status.
Power integrity, sensor cable strain relief, a common ground, ADC reference stability, and wearable comfort are as important as the component list. Bluetooth command mode and data mode should not be confused during integration.
Software architecture
Separate real-time tasks can sample finger sensors, read inertial data, filter and calibrate measurements, update the display, and transmit output. Shared data needs a clear synchronization strategy; task priorities and periods should be based on measured timing rather than arbitrary values.
A transmitted frame should include a start marker, version or type, payload length, sequence number, calibrated finger values, orientation data, and checksum. The receiver must reject malformed or stale frames and move the controlled system to a safe state after communication loss.
Calibration and applications
Calibration records the usable range for each finger and maps it to a normalized value. Filtering reduces ADC noise, while orientation fusion limits accelerometer jitter and gyroscope drift. Test static poses, slow and rapid motion, repeated donning, radio interruption, and battery variation.
The prototype can map the wearer’s hand to a virtual hand or command a five-axis robotic arm. A physical robot additionally requires joint limits, speed limits, an emergency stop, and independent safety logic; wearable input should never be the only protection against hazardous motion.
