Page 269 - Emerging Trends and Innovations in Web-Based Applications and Technologies

International Journal of Trend in Scientific Research and Development (IJTSRD) @ www.ijtsrd.com eISSN: 2456-6470
Hypothetical Example of Results REFERENCES
For clarity, we expect the following performance results for [1] Koller, O, Zargaran, S., Ney, H.; Bowden, R. Deep sign:
the ISL dataset: Enabling robust statistical continuous sign language
recognition via hybrid CNN-HMMs. Int. J. Comput. Vis
Expected
Text to Sign Precision Recall 2018, 126, 1311-1325.
accuracy
“Hello” 95% 87% 88% [2] Ananthanarayana, T., Srivastava, P. Chintha, A.;
Santha, A.; Landy, B.; Panaro, J., Webster, A., Kotecha,
“Thank You” 98% 90% 91% N., Sah, S.; Sarchet, T.; et al. Deep learning methods for
“Sorry” 87% 92% 84% sign language translation ACM Trans, Access. Comput.
(TACCESS) 2021, 14,1-30.
“Yes” 91% 95% 90%
[3] Koller, O; Ney, H.; Bowden, R. Deep hand: How to train
“No” 78% 80% 73% a cin on 1 million hand images when your data is
“Please” 87% 88% 85% continuous and weakly labelled. In Proceedings of the
IEEE Conference on Computer Vision and Pattern
“Help” 84% 85% 81% Recognition, Las Vegas, NV, USA, 27-30 June 2016, pp.
“Good Morning” 82% 90% 78% 3793-3802.
“Goodbye” 88% 85% 87% [4] Kumar, S.S., Wangyal, T.; Saboo, V., Srinath, R. Time
series neural networks for real time sign language
“Occluded Gestures” 75% 72% 73% translation. In Proceedings of the 2018 17th IEEE
International Conference on Machine Learning and
The CNN-based model for ISL recognition is expected to Applications (ICMLA), Orlando, FL, USA, 17-20
achieve high accuracy, fast processing times, and robust December 2018; pp. 243-248.
performance across various gesture categories. We
anticipate that the model will be a valuable tool for [5] Rokade, Y. I., & Jadav, P. M. (2017), "Indian Sign
facilitating communication in Indian Sign Language, Language Recognition System", International Journal
particularly for real-time applications such as sign language of Engineering and Technology (IJET), Volume 9, Issue
interpretation and human-computer interaction. 3S, PP. 189-195,
https://www.researchgate.net/publication/3186569
V. CONCLUSION 56_Indian_Sign_Language_Recognition_System
In this study, we explored the effectiveness of Convolutional
Neural Networks (CNNs) in the detection and translation of [6] Zhou, H., Zhou, W., Zhou, Y, LI, H. Spatial-temporal
sign language gestures. The results demonstrated a multi-cue network for sign language recognition and
promising accuracy rate of 92% in recognizing Indian Sign translation. IEEE Trans. Multimed 2021, 24, 768-779
Language (ISL) gestures using a CNN-based model trained on [7] Moryossef, A.; Yin, K.; Neubig, G., Goldberg, Y. Data
a diverse dataset of hand shapes and motions. This high
accuracy highlights the capability of CNNs in processing augmentation for sign language gloss translation,
spatial features and distinguishing intricate gesture patterns, arXiv 2021, arXiv: 2105.07476.
even under varying environmental conditions. The real-time [8] M. F., K., R. F., & V. S., A. (2024), "Malayalam Sign
performance of the system showed a detection speed of 98%, Language Identification using Finetuned YOLOv8 and
with minimal latency, confirming the feasibility of applying Computer Vision Techniques", arXiv preprint
this model in live scenarios. arXiv:2405.06702, https://arxiv.org/abs/2405.06702
However, the study also acknowledged certain challenges [9] Cao, Y.; Li, W.; LI, X. Chen, M., Chen, G; Hu, L.; LI, Z; Kai,
that could affect accuracy, such as lighting conditions, hand H. Explore more guidance: A task-aware instruction
occlusion, and varying gesture speeds. Moreover, the network for sign language translation enhanced with
accuracy of text-to-sign systems, which is currently around data augmentation.
75%, will benefit from further enhancements in 3D gesture
modeling and real-time translation capabilities. The use of [10] Verma, R., Gupta, S., & Arora, P. (2022), "Real-time
CNNs for sign language detection shows significant potential Gesture Recognition for Indian Sign Language using
in facilitating communication between hearing and non- YOLOv5", (https://arxiv.org/abs/2205.09834).
hearing individuals. With continued advancements in model [11] Patel, P., Shah, R., & Mehta, K. (2021), "Gesture-based
robustness and dataset expansion, sign language detection Communication Systems for the Hearing Impaired: A
systems could become more accurate, reliable, and scalable, Review", Journal of Human-Computer Interaction,
ultimately supporting greater inclusivity across different Volume 32, Issue 4, PP. 289-302, (https://hci-
social, educational, and professional environments. journal.com/2021/04/gesture-systems).
VI. FUTURE SCOPE [12] Kumar & Singh, R. (2019), "A Comprehensive Survey
Sign language detection offers vast potential to improve on Sign Language Recognition Systems", International
communication and inclusivity in areas like education, Journal of Advanced Research in Computer Science,
healthcare, workplaces, and entertainment. By utilizing AI, Volume 10, Issue 3, PP. 45-
AR, and IoT, it enables real-time sign-to-speech and speech- 52(https://doi.org/10.1016/ijarcs.2019.03.005).
to-sign translation, enhancing accessibility in public spaces
and virtual platforms. Future developments aim to support [13] Ramesh, S., & Das, T. (2020), "Deep Learning
diverse sign languages, address regional differences, and Approaches for Indian Sign Language Recognition",
improve scalability and accuracy, fostering seamless IEEE Transactions on Artificial Intelligence, PP. 123-
communication and inclusivity.

IJTSRD | Special Issue on Emerging Trends and Innovations in Web-Based Applications and Technologies Page 259