Breaking Barriers: Real-Time Sign Language Recognition Using LSTM Networks for Enhanced Communication Accessibility
Bibliographic record
Abstract
This paper presents an innovative approach to real-time sign language recognition using Long Short-Term Memory networks (LSTM), aimed at enhancing communication accessibil-ity for the deaf and hard-of-hearing community. We address the challenge of understanding and interpreting sign language, which is critical for millions worldwide, yet restricted to those proficient in it. Our research contributes to bridging this communication gap by developing a deep learning model capable of recognizing a broad spectrum of sign language gestures and sentences with high accuracy and speed. Utilizing a rich dataset comprising diverse sign language gestures, collected in collaboration with a professional video production studio and proficient sign language users, we employ LSTM networks integrated with Dense layers to effectively capture the complex spatial and temporal patterns of sign language. The architecture of our model is specifically designed to accommodate the nuanced dynamics of sign language, with an emphasis on real-time processing. Through rigorous training and validation, our model demonstrates an outstanding accuracy rate of 92 % on a comprehensive testing dataset, alongside remarkable real-time processing capabilities. The sys-tem's efficiency in recognizing a wide array of sign gestures nearly instantaneously underscores its potential applicability in various real-world scenarios, including assistive technologies and human-computer interaction. This study not only showcases the practicality and efficacy of LSTM networks in real-time sign language recognition but also marks a significant step towards more inclusive and accessible communication technologies. Our future work includes integrating this system with the Langue des Signes Quebecoise website, further advancing the goal of universal communication accessibility.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".