Scholarlink

<p>Autism Spectrum Disorder (ASD) continues to pose a significant healthcare concern and attract great attention owing to rising incidence rates and the substantial cost on families and society. Emotion recognition is critical in the therapeutic and developmental support of children with intense-level autism, particularly for those in hospital settings where behavioral cues can be subtle or challenging to interpret. This study proposes the SpAuCNN (Sparse Autoencoder Convolutional Neural Network), a novel architecture designed to accurately detect and interpret emotions in children with intense autism spectrum disorder (ASD). It combines sparse autoencoder networks with convolutional neural networks (CNN), aiming to enhance feature extraction by VGG-19 (visual geometry group), which can isolate and preserve high-dimensional emotional cues often obscured by noise and variability in facial expressions and body language. To evaluate the effectiveness of SpAuCNN, we collected a specialized dataset consisting of annotated facial and EEG data from hospitalized children diagnosed with intense-level ASD. The experimental results demonstrate that SpAuCNN achieves 98.34% of accuracy, 97.56% of precision, 96.53% of recall and 97.51% of f1-score.</p>

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

SpAuCNN: Sparse autoencoder convolution neural network architecture for emotion recognition among intense level autism disorder hospitalized children

  • 期刊论文

  • 发表日期：

    2025年10月25日

查看全文
  • Veerendra Bethineedi,
  • Ramakrishna Damodar

摘要

Autism Spectrum Disorder (ASD) continues to pose a significant healthcare concern and attract great attention owing to rising incidence rates and the substantial cost on families and society. Emotion recognition is critical in the therapeutic and developmental support of children with intense-level autism, particularly for those in hospital settings where behavioral cues can be subtle or challenging to interpret. This study proposes the SpAuCNN (Sparse Autoencoder Convolutional Neural Network), a novel architecture designed to accurately detect and interpret emotions in children with intense autism spectrum disorder (ASD). It combines sparse autoencoder networks with convolutional neural networks (CNN), aiming to enhance feature extraction by VGG-19 (visual geometry group), which can isolate and preserve high-dimensional emotional cues often obscured by noise and variability in facial expressions and body language. To evaluate the effectiveness of SpAuCNN, we collected a specialized dataset consisting of annotated facial and EEG data from hospitalized children diagnosed with intense-level ASD. The experimental results demonstrate that SpAuCNN achieves 98.34% of accuracy, 97.56% of precision, 96.53% of recall and 97.51% of f1-score.