Hand sign language recognition using multi-view hand skeleton

作者：

Highlights：

• We propose a model for hand sign recognition using SSD, 3DCNN, and LSTM from RGB.

• We propose a dataset including 10′000 RGB sign videos.

• We build hand skeleton using multi-view projection of 3D hand keypoints.

• Our model outperforms state-of-the-art models on NYU and First-Person datasets.

• We apply 3DCNN on stacked inputs to get discriminant local spatio-temporal features.

摘要

•We propose a model for hand sign recognition using SSD, 3DCNN, and LSTM from RGB.•We propose a dataset including 10′000 RGB sign videos.•We build hand skeleton using multi-view projection of 3D hand keypoints.•Our model outperforms state-of-the-art models on NYU and First-Person datasets.•We apply 3DCNN on stacked inputs to get discriminant local spatio-temporal features.

论文关键词：Multi-view hand skeleton,Hand sign language recognition,3DCNN,Hand pose estimation,RGB video,Hand action recognition

论文评审过程：Received 18 September 2019, Revised 26 December 2019, Accepted 21 February 2020, Available online 22 February 2020, Version of Record 11 March 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113336