Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Handwriting Anomalies and Learning Disabilities through Recurrent Neural Networks and Geometric Pattern Analysis

Vasileios Alevizos, Sabrina Edralin, Akebu Simasiku, Dimitra Malliarou, Antonis Messinis, George Papakostas, Clark Xu, Zongliang Yue

TL;DR

This work tackles automatic detection of dyslexia and dysgraphia from handwriting by modeling geometric and temporal writing patterns. It proposes a hybrid pipeline that extracts features such as baseline deviation, letter connectivity, and stroke thickness, then processes them with an RNN-based autoencoder to identify irregularities. The system is trained on a diverse dataset of about 33,000 handwriting samples across normal, dyslexia, and dysgraphia categories, achieving state-of-the-art performance for combined detection and revealing challenges in generalizing to diverse corpora. The work highlights the importance of sequential modeling for handwriting and points to future improvements via CNNs and self-attention mechanisms to better handle long-range dependencies and multi-modal data.

Abstract

Dyslexia and dysgraphia are learning disabilities that profoundly impact reading, writing, and language processing capabilities. Dyslexia primarily affects reading, manifesting as difficulties in word recognition and phonological processing, where individuals struggle to connect sounds with their corresponding letters. Dysgraphia, on the other hand, affects writing skills, resulting in difficulties with letter formation, spacing, and alignment. The coexistence of dyslexia and dysgraphia complicates diagnosis, requiring a nuanced approach capable of adapting to these complexities while accurately identifying and differentiating between the disorders. This study utilizes advanced geometrical patterns and recurrent neural networks (RNN) to identify handwriting anomalies indicative of dyslexia and dysgraphia. Handwriting is first standardized, followed by feature extraction that focuses on baseline deviations, letter connectivity, stroke thickness, and other anomalies. These features are then fed into an RNN-based autoencoder to identify irregularities. Initial results demonstrate the ability of this RNN model to achieve state-of-art performance on combined dyslexia and dysgraphia detection, while showing the challenges associated with complex pattern adaptation of deep-learning to a diverse corpus of about 33,000 writing samples.

Handwriting Anomalies and Learning Disabilities through Recurrent Neural Networks and Geometric Pattern Analysis

TL;DR

This work tackles automatic detection of dyslexia and dysgraphia from handwriting by modeling geometric and temporal writing patterns. It proposes a hybrid pipeline that extracts features such as baseline deviation, letter connectivity, and stroke thickness, then processes them with an RNN-based autoencoder to identify irregularities. The system is trained on a diverse dataset of about 33,000 handwriting samples across normal, dyslexia, and dysgraphia categories, achieving state-of-the-art performance for combined detection and revealing challenges in generalizing to diverse corpora. The work highlights the importance of sequential modeling for handwriting and points to future improvements via CNNs and self-attention mechanisms to better handle long-range dependencies and multi-modal data.

Abstract

Dyslexia and dysgraphia are learning disabilities that profoundly impact reading, writing, and language processing capabilities. Dyslexia primarily affects reading, manifesting as difficulties in word recognition and phonological processing, where individuals struggle to connect sounds with their corresponding letters. Dysgraphia, on the other hand, affects writing skills, resulting in difficulties with letter formation, spacing, and alignment. The coexistence of dyslexia and dysgraphia complicates diagnosis, requiring a nuanced approach capable of adapting to these complexities while accurately identifying and differentiating between the disorders. This study utilizes advanced geometrical patterns and recurrent neural networks (RNN) to identify handwriting anomalies indicative of dyslexia and dysgraphia. Handwriting is first standardized, followed by feature extraction that focuses on baseline deviations, letter connectivity, stroke thickness, and other anomalies. These features are then fed into an RNN-based autoencoder to identify irregularities. Initial results demonstrate the ability of this RNN model to achieve state-of-art performance on combined dyslexia and dysgraphia detection, while showing the challenges associated with complex pattern adaptation of deep-learning to a diverse corpus of about 33,000 writing samples.
Paper Structure (12 sections, 4 figures)

This paper contains 12 sections, 4 figures.

Table of Contents

  1. Introduction
  2. Poor Handwriting Association with Learning Disabilities
  3. Visual Diagnosis of Dyslexia and Dysgraphia
  4. Contributions of Artificial Systems in Detection
  5. METHODOLOGY
  6. Dataset
  7. Building Blocks and Ingredients
  8. Process of Text Irregularities
  9. Evaluation
  10. Results
  11. DISCUSSION AND FUTURE DIRECTIONS
  12. CONCLUSION

Figures (4)

  • Figure 1: Workflow of handwriting anomalies detection starting with an inspection of relevant features that will later be used in the upcoming steps of anomaly detection.
  • Figure 2: Computational graph visualization of a Recurrent Neural Network during back propagation.
  • Figure 3: Analysis of training dynamics showing loss decrease and accuracy increase over 120 epochs, rapid improvement is present at the beginning of training followed by stabilization as accuracy sharply increases.
  • Figure 4: Comparative analysis of model accuracy across different handwriting conditions over epochs. Initially, accuracy increases across all categories. Dyslexia and ground truth adapted fastest; low potential dysgraphia adapted slowest; normal handwriting performed best overall.