Handwritten Sanskrit Word Recognition: A Deep Learning Approach Using AlexNet

Year : 2025 | Volume : 15 | Issue : 02 | Page : 33 43
    By

    Shraddha V. Shelke,

  • D.M. Chandwadkar,

  • S.P. Ugale,

  • Rupali V. Chothe,

  1. Research Scholar, Department of Electronics and Telecommunication, Karmaveer Kakasaheb Wagh Institute of Engineering Education & Research, Nashik, Savitribai Phule Pune University, Pune, Maharashtra, India
  2. Professor and Head, Department of Electronics and Telecommunication, Karmaveer Kakasaheb Wagh Institute of Engineering Education & Research, Nashik, Savitribai Phule Pune University, Pune, Maharashtra, India
  3. Professor, Department of Electronics and Telecommunication, Karmaveer Kakasaheb Wagh Institute of Engineering Education & Research (KKWIEE & R), Nashik, Maharashtra, India
  4. Assistant Professor, Department of Electronics and Telecommunication, Karmaveer Kakasaheb Wagh Institute of Engineering Education & Research (KKWIEE & R), Nashik, Maharashtra, India

Abstract

Handwritten Sanskrit word recognition poses significant challenges due to the intricate structure of the script and the considerable variations in handwriting across individuals. To address these challenges, this research introduces a novel methodology employing transfer learning with the AlexNet convolutional neural network. The study utilized two distinct datasets: a specifically curated Sanskrit word image dataset containing 2616 samples, alongside a broader Devanagari character dataset used for validation purposes. The established AlexNet architecture underwent a process of fine-tuning, primarily focused on the Sanskrit word dataset. Key hyperparameters, such as the learning rate and the number of training epochs, were carefully optimized to maximize the model’s performance. The proposed approach demonstrated promising outcomes, achieving results that surpassed existing state- of-the-art techniques on both datasets. Notably, with a training duration of 20 epochs and an optimized learning rate of 0.015, the model achieved a peak testing accuracy of 97.9%. To enhance the model’s training efficiency, the Stochastic Gradient Descent with Momentum (SGDM) optimizer was strategically implemented, effectively accelerating the convergence process and mitigating oscillations during learning. This research highlights the effectiveness of leveraging transfer learning with deep convolutional neural networks for the complex task of handwritten Sanskrit word recognition.

Keywords: Sanskrit word recognition, Alexnet, convolutional neural network, stochastic gradient descent with Momentum

[This article belongs to Current Trends in Signal Processing ]

How to cite this article:
Shraddha V. Shelke, D.M. Chandwadkar, S.P. Ugale, Rupali V. Chothe. Handwritten Sanskrit Word Recognition: A Deep Learning Approach Using AlexNet. Current Trends in Signal Processing. 2025; 15(02):33-43.
How to cite this URL:
Shraddha V. Shelke, D.M. Chandwadkar, S.P. Ugale, Rupali V. Chothe. Handwritten Sanskrit Word Recognition: A Deep Learning Approach Using AlexNet. Current Trends in Signal Processing. 2025; 15(02):33-43. Available from: https://journals.stmjournals.com/ctsp/article=2025/view=215236


References

  1. Moudgil A, Singh S, Gautam V, Rani S, Shah SH. Handwritten Devanagari manuscript characters recognition using capsnet. International Journal of Cognitive Computing in Engineering. 2023 Jun 1; 4: 47-–54.
  2. More V, Kharat M, Gumaste S. Segmentation of devanagari handwritten text using thresholding approach. Int. J. Sci. Technol. Res.(IJSTR). 2020 Mar; 9(3): 2277-–8616.
  3. Gurav Y, Bhagat P, Jadhav R, Sinha S. Devanagari handwritten character recognition using convolutional neural networks. In 2020 IEEE International Conference on Electrical, Communication, and Computer Engineering (ICECCE). 2020 Jun 12 (pp.; 1-–6). IEEE.
  4. Shelke SV, Chandwadkar DM, Ugale SP, Chothe RV. Combining Multiple Feature Extraction and Classification Methods to Study Performance of Handwritten Sanskrit Character Recognition. In 2023 IEEE 7th International Conference On Computing, Communication, Control And Automation (ICCUBEA). 2023 Aug 18 (pp.; 1-–6). IEEE.
  5. Lomte MV, Doye DD. Handwritten Vedic Sanskrit text recognition using deep learning. Journal of Algebraic Statistics. 2022 Jul 12; 13(3): 2190-–8.
  6. Aneja N, Aneja S. Transfer learning using CNN for handwritten Devanagari character recognition. In 2019 IEEE 1st International Conference on Advances in Information Technology (ICAIT). 2019 Jul 25 (pp.; 293-–296). IEEE.
  7. Gupta P, Deshmukh S, Pandey S, Tonge K, Urkunde V, Kide S. Convolutional neural network based handwritten Devanagari character recognition. In 2020 IEEE International conference on smart technologies in computing, electrical and electronics (ICSTCEE). 2020 Oct 9 (pp.; 322-–326). IEEE.
  8. Deore SP, Pravin A. Devanagari handwritten character recognition using fine-tuned deep convolutional neural network on trivial dataset. Sādhanā. 2020 Dec; 45(1): 243.
  9. Shingate MD, Gumaste SV. Improving Ensemble Classifier for Natural Language Situational Information Analysis. In 2024 IEEE International Conference on Emerging Innovations and Advanced Computing (INNOCOMP). 2024 May 25 (pp.; 305-–311). IEEE.
  10. Shelke SV, Chandwadkar DM. Automatic system for recognition of handwritten character using multiscale neural network. In: Proceedings of the International Conference in Computational Intelligence (ICCIA); 2012 Mar;6:16–206.
  11. Bisht M, Gupta R. Multiclass recognition of offline handwritten Devanagari characters using CNN. International Journal of Mathematical, Engineering and Management Sciences. 2020; 5(6): 1429-–39.
  12. Avadesh M, Goyal N. Optical character recognition for Sanskrit using convolution neural networks. In 2018 13th IAPR international workshop on document analysis systems (DAS). 2018 Apr 24 (pp.; 447-–452). IEEE.
Regular Issue Subscription Original Research
Volume 15
Issue 02
Received 14/05/2025
Accepted 16/05/2025
Published 25/06/2025
Publication Time 42 Days
117

Login


My IP

PlumX Metrics