Scholarlink

Speech abnormalities affect millions globally, particularly in underserved areas with limited access to speech-language pathologists. Traditional diagnostic methods are labor-intensive and subjective, necessitating automated solutions. This study explores machine learning for speech disorder classification using audio data. Two methods were tested: (1) CNNs with Mel Frequency Spectrograms, achieving 99.87% accuracy, and (2) traditional models with Mel Frequency Cepstral Coefficients (MFCC) and Support Vector Machines (SVM), achieving 99.76% accuracy followed by random forest and KNN with an accuracy of 99.523%. The results demonstrate how machine learning can enhance early detection and accessibility of speech disorder diagnosis, especially in underprivileged regions.

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

Speech Disorder Detection Using Convolution Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Shouryya Veer Monga,
  • Ritik Kumar Chaudhary,
  • Deepika Kumar,
  • Jahanavi,
  • Vani Jain,
  • Alaa Ali Hameed

摘要

Speech abnormalities affect millions globally, particularly in underserved areas with limited access to speech-language pathologists. Traditional diagnostic methods are labor-intensive and subjective, necessitating automated solutions. This study explores machine learning for speech disorder classification using audio data. Two methods were tested: (1) CNNs with Mel Frequency Spectrograms, achieving 99.87% accuracy, and (2) traditional models with Mel Frequency Cepstral Coefficients (MFCC) and Support Vector Machines (SVM), achieving 99.76% accuracy followed by random forest and KNN with an accuracy of 99.523%. The results demonstrate how machine learning can enhance early detection and accessibility of speech disorder diagnosis, especially in underprivileged regions.