Scholarlink

In this paper, a speech-based detection system is developed for voice pathologies like GERD, laryngitis, bulbar palsy, vocal fold leukoplakia, cyst, dysarthria, and aphasia. The e-health set-up comprises three modules for the detection of specific voice disorders leading to degradation in voice quality. Classification of these pathologies is performed using various machine learning algorithms before and after data balancing using SMOTE. The maximum performance is obtained for CatBoost with an accuracy of 56% for the detection of voice disorders before data balancing and 93% post-data balancing. An accuracy of 86.5% is obtained by random forest for detection of dysarthria and an accuracy of 88% is obtained for aphasia using the CatBoost classifier. Further, to make an improvement in the selected features we have customized a generalized feature set for the classification of voice disorders. All the components are effectively incorporated into a user-friendly e-health package, ensuring non-invasive, and precise diagnosis.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design and Development of an e-Health Package for Voice Pathology Detection

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shiksha Rai,
  • S. Rishita,
  • Kandalam Srihari Lasya,
  • Susmitha Vekkot

摘要

In this paper, a speech-based detection system is developed for voice pathologies like GERD, laryngitis, bulbar palsy, vocal fold leukoplakia, cyst, dysarthria, and aphasia. The e-health set-up comprises three modules for the detection of specific voice disorders leading to degradation in voice quality. Classification of these pathologies is performed using various machine learning algorithms before and after data balancing using SMOTE. The maximum performance is obtained for CatBoost with an accuracy of 56% for the detection of voice disorders before data balancing and 93% post-data balancing. An accuracy of 86.5% is obtained by random forest for detection of dysarthria and an accuracy of 88% is obtained for aphasia using the CatBoost classifier. Further, to make an improvement in the selected features we have customized a generalized feature set for the classification of voice disorders. All the components are effectively incorporated into a user-friendly e-health package, ensuring non-invasive, and precise diagnosis.