Scholarlink

In this paper, we investigate the performance of the vision-based American Sign Language (ASL) recognition system employing the Long Short-Term Memory (LSTM) classifier. MediaPipe introduced by Google [15] is used to obtain posture landmarks where a custom dataset has been created and used for the experimental study. Ten sign words were considered from the ASL: hello, iloveyou, thanks, google, internet, jogging, house, book, drink and drive. The proposed system is benchmarked with the classifier employing the k Nearest Neighbor (kNN) with Dynamic Time Warping (DTW). The proposed system yields 92% accuracy comparable to those obtained from the classifier using the kNN with DTW whose accuracy is equal to 75%. In addition, we evaluated the robustness of the proposed system by corrupting the landmarks with zero-mean additive white Gaussian noise whose standard deviation (SD) ranges from 0.001 to 1. The proposed system maintains accuracy beyond 90% when noise SD is less than or equal to 0.15 but exhibits significant performance drop when the noise SD is beyond this value.

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

Performance of the Sign Language Recognition System Using the Long Short-Term Memory Network Algorithm

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Lakash Maharjan,
  • Chantri Polprasert

摘要

In this paper, we investigate the performance of the vision-based American Sign Language (ASL) recognition system employing the Long Short-Term Memory (LSTM) classifier. MediaPipe introduced by Google [15] is used to obtain posture landmarks where a custom dataset has been created and used for the experimental study. Ten sign words were considered from the ASL: hello, iloveyou, thanks, google, internet, jogging, house, book, drink and drive. The proposed system is benchmarked with the classifier employing the k Nearest Neighbor (kNN) with Dynamic Time Warping (DTW). The proposed system yields 92% accuracy comparable to those obtained from the classifier using the kNN with DTW whose accuracy is equal to 75%. In addition, we evaluated the robustness of the proposed system by corrupting the landmarks with zero-mean additive white Gaussian noise whose standard deviation (SD) ranges from 0.001 to 1. The proposed system maintains accuracy beyond 90% when noise SD is less than or equal to 0.15 but exhibits significant performance drop when the noise SD is beyond this value.