Document Type : Original Article

Authors

• سید علی زنده باد ¹
• Fatemeh Darkhal ²
• Zahra Sedaghat ³

¹ Department of Biomedical Engineering Mashhad Branch, Islamic Azad University Mashhad, Iran.

² Department of Electrical and Biomedical Engineering University College of Rouzbahan Sari, Iran.

³ Clinical Research Development Unit of Bou Ali Sina Hospital School of Medicine, Mazandaran University of Medical Sciences Sari, Iran

Abstract

Abstract—Abstract—Early diagnosis of Parkinson's disease (PD) is an important challenge in medicine. Hand tremors and writing disorders, which are early motor symptoms of Parkinson's, would appear before a formal diagnosis for decades. So, handwriting analysis has become an essential tool for diagnosing Parkinson's disease. While many machine learning algorithms have been applied in this field, they struggle to capture subtle variations in handwriting and must describe features from different perspectives. To address these problems, this study proposes a model for Parkinson's handwriting recognition. This long-term dependence of the features on the common coordinate attention schedule enables the model to more accurately localize important features of handwriting data and extract fuzzy edge features of handwriting images. These features of the CAS transformer will allow it to outperform current state-of-the-art deep learning methods in classification, with an accuracy of 92.68% in experiments conducted on two handwritten datasets. Abstract— Abstract— Abstract— Abstract— Abstract—

Keywords

• Keywords: Data Augmentation, Deep Learning (DL), Handwriting Analysis, Parkinson'
• s Disease (PD), Pattern Recognition, Transformer

Main Subjects

• Artificial Intelligence, Robotics, Cognitive Computing