Scholarlink

In computer vision, dynamic hand gesture detection is an attractive and difficult area. Determining when a gesture begins and finishes in a video is one of the primary challenges, particularly when the gestures are continuous. Because the system must comprehend both spatio-temporal patterns, this makes real-time gesture recognition challenging. This is becoming easier because of new techniques like S3D, which captures specific movement details, and LSTM, which analyzes patterns across time. Using these methods, hand gesture detection becomes more accurate and flexible for a variety of users and conditions. In this study, we extracted features from hand gestures using the S3D model. These characteristics were then fed into an LSTM model for hand gesture classification. Hand detection and tracking problems are successfully addressed by feature extraction using the S3D model. Three benchmark datasets FPHA, SKIG, and CHG, were used to thoroughly test our approach, and the results showed remarkable accuracy rates of \(97.3\%\) , \(99.2\%\) , and \(97.8\%\) , respectively. These findings show that our suggested model works better than current state-of-the-art methods, providing a very practical answer for real-time gesture detection applications.

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

A Deep Learning Approach to Dynamic Hand Gesture Recognition with S3D, 1D-CNN, and LSTM

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Reena Tripathi,
  • Bindu Verma

摘要

In computer vision, dynamic hand gesture detection is an attractive and difficult area. Determining when a gesture begins and finishes in a video is one of the primary challenges, particularly when the gestures are continuous. Because the system must comprehend both spatio-temporal patterns, this makes real-time gesture recognition challenging. This is becoming easier because of new techniques like S3D, which captures specific movement details, and LSTM, which analyzes patterns across time. Using these methods, hand gesture detection becomes more accurate and flexible for a variety of users and conditions. In this study, we extracted features from hand gestures using the S3D model. These characteristics were then fed into an LSTM model for hand gesture classification. Hand detection and tracking problems are successfully addressed by feature extraction using the S3D model. Three benchmark datasets FPHA, SKIG, and CHG, were used to thoroughly test our approach, and the results showed remarkable accuracy rates of \(97.3\%\) , \(99.2\%\) , and \(97.8\%\) , respectively. These findings show that our suggested model works better than current state-of-the-art methods, providing a very practical answer for real-time gesture detection applications.