Scholarlink

<p>Hand gesture recognition (HGR) is critical in prosthetics, rehabilitation, and human–machine interaction. This study introduces a quantum-inspired neural framework for multimodal HGR using surface electromyography (sEMG) and CyberGlove signals from the NINAPRO DB1 dataset. The dataset includes 52 gestures performed by 10 subjects, captured with 10 sEMG electrodes and 22 CyberGlove joint- angle sensors. Preprocessing involved segmentation, statistical feature extraction, and dimensionality reduction via principal component analysis (PCA). The features were mapped into quantum-inspired represen- tations and classified using three architectures: Quantum-Inspired Recurrent Neural Network (Q-RNN), Quantum-Inspired Convolutional Neural Network (Q-CNN), and Quantum Ensemble Design (QED). Experimental results showed that QED achieved the highest performance with 77.5% accuracy, outperform- ing both Q- RNN (70.0%) and Q-CNN (50.3%). These findings demonstrate the effectiveness of ensemble based quantum models for multimodal gesture recog- nition and highlight their potential for real-time applications in next generation human–computer interaction systems.</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

Hand Gesture Recognition Using Quantum Inspired Neural Networks with sEMG

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Giftlin Olivia,
  • Pramod Mathew Jacob,
  • Thejaswini Kishore,
  • M. S. P. Subathra

摘要

Hand gesture recognition (HGR) is critical in prosthetics, rehabilitation, and human–machine interaction. This study introduces a quantum-inspired neural framework for multimodal HGR using surface electromyography (sEMG) and CyberGlove signals from the NINAPRO DB1 dataset. The dataset includes 52 gestures performed by 10 subjects, captured with 10 sEMG electrodes and 22 CyberGlove joint- angle sensors. Preprocessing involved segmentation, statistical feature extraction, and dimensionality reduction via principal component analysis (PCA). The features were mapped into quantum-inspired represen- tations and classified using three architectures: Quantum-Inspired Recurrent Neural Network (Q-RNN), Quantum-Inspired Convolutional Neural Network (Q-CNN), and Quantum Ensemble Design (QED). Experimental results showed that QED achieved the highest performance with 77.5% accuracy, outperform- ing both Q- RNN (70.0%) and Q-CNN (50.3%). These findings demonstrate the effectiveness of ensemble based quantum models for multimodal gesture recog- nition and highlight their potential for real-time applications in next generation human–computer interaction systems.