I designed and evaluated the background-elimination preprocessing method, then helped integrate it with the transfer-learning and random-forest recognition pipeline.
COMPLETED RESEARCH CASE STUDYMULTI-AUTHOR WORK
Bangla Sign Language Recognition
Undergraduate thesis research that contributed to a published hybrid recognition system combining background elimination, transfer learning, and a random-forest classifier.
MY CONTRIBUTION
Preprocessing design and pipeline integration.
The full system and reported metrics belong to the multi-author publication; this case study does not present the research as solo work.
SYSTEM METHOD
A hybrid path from sign image to classification.
The published method separates background removal, learned feature extraction, and final classification.
Sign image
Open Bangla Sign Language datasets provide character and digit inputs.
Background elimination
Preprocessing removes unwanted visual features.
Transfer learning
Pre-trained backbone networks extract useful representations.
Random forest
The classifier produces the recognition result.
Read the published method and results.
The two accuracy figures above are paper-level results from the multi-author system on Ishara-Lipi and Ishara-Bochon.
Open the paper