All projects
03 / Jan 2023/ KCG College of Technology
Sign Language Interpreting Smart Eyewear
A person who signs cannot easily talk with someone who does not know sign language, and glove-based interpreters are awkward to wear.
- Role
- Vision pipeline and wearable hardware design.
- Stack
- Python, OpenCV, MediaPipe, Raspberry Pi 4B, CNN
- Patent
- Indian patent application filed Feb 10, 2023
01
What I built
- 01Glasses with a built-in camera that read Indian Sign Language gestures and speak them aloud in real time.
- 02Fully on-device on embedded Linux: 5 MP CSI camera, OpenCV preprocessing, MediaPipe 21-point hand landmarks, a CNN gesture classifier, then text-to-speech. No cloud connection.
- 03Camera only, so the user wears nothing on their hands.
02
Hardware
- 01Raspberry Pi 4B (ARM Cortex-A72)
- 025 MP CSI camera, speaker
- 033.7 V 3250 mAh Li-Po with LM2596 buck regulation
- 04Head-worn packaging with cooling
03
Software
- 01Python on embedded Linux
- 02OpenCV
- 03MediaPipe hand landmarks
- 04CNN classifier
- 05Text-to-speech
04
Validation
- 01Real-time gesture-to-speech pipeline demonstrated on the wearable prototype.
05
Results
- 01Indian patent application filed February 10, 2023.
06
Characteristics
| Parameter | Value |
|---|---|
| Compute | Raspberry Pi 4B, Cortex-A72 |
| Camera | 5 MP CSI |
| Hand model | MediaPipe, 21 landmarks |
| Battery | 3.7 V, 3250 mAh Li-Po |
| Regulation | LM2596 buck |
| Connectivity needed | None, on-device |
Links
07
Gallery
