Scholarlink

Spasmodic Dysphonia is a neurological disorder that mainly affects the voice muscles. This disease can touch the adductor muscles, the abductor muscles, or both. The involuntary muscle spasms make speech difficult and can be very disabling. Diagnosis and assessment of the patient's condition often involve invasive and costly procedures such as fibroscopy or magnetic resonance imaging. Acoustic analysis of the voice represents an affordable and non-invasive alternative. In this perspective, Clinicians can be assisted by artificial intelligence to perform preliminary, rapid, and less costly diagnoses. Machine learning remains a highly promising field and has yielded excellent results. In this paper, we have constructed a system for detecting spasmodic dysphonia (SD) from voice recordings. The paper proposes a 1D Convolutional Neural Network (1D-CNN) for detecting spasmodic dysphonia (SD) using voice recordings from the Saarbrucken Voice Database. The model, enhanced by mRMR feature selection, achieves 93.42% accuracy and an F1-score of 0.93. The approach offers a non-invasive, cost-effective diagnostic tool, with potential for embedded system implementation.

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

High-Accuracy Spasmodic Dysphonia Detection Using Machine Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Hassan Ezzahori,
  • Abdelkrim Hammimou,
  • Abdelghani Boudaoud,
  • Mounaim Aqil

摘要

Spasmodic Dysphonia is a neurological disorder that mainly affects the voice muscles. This disease can touch the adductor muscles, the abductor muscles, or both. The involuntary muscle spasms make speech difficult and can be very disabling. Diagnosis and assessment of the patient's condition often involve invasive and costly procedures such as fibroscopy or magnetic resonance imaging. Acoustic analysis of the voice represents an affordable and non-invasive alternative. In this perspective, Clinicians can be assisted by artificial intelligence to perform preliminary, rapid, and less costly diagnoses. Machine learning remains a highly promising field and has yielded excellent results. In this paper, we have constructed a system for detecting spasmodic dysphonia (SD) from voice recordings. The paper proposes a 1D Convolutional Neural Network (1D-CNN) for detecting spasmodic dysphonia (SD) using voice recordings from the Saarbrucken Voice Database. The model, enhanced by mRMR feature selection, achieves 93.42% accuracy and an F1-score of 0.93. The approach offers a non-invasive, cost-effective diagnostic tool, with potential for embedded system implementation.