In-air handwritten Chinese text recognition with temporal convolutional recurrent network

作者：

Highlights：

• We present a 3D touchless in-air handwritten Chinese text recognition system for the first time.

• We constructed the first publicly available in-air handwritten Chinese text dataset, named IAHCT-UCAS2018, for the community.

• A novel architecture, named the temporal convolutional recurrent network, is proposed for online handwritten Chinese text recognition, which achieves the competitive performance compared with the state-of-the-art methods on CASIA-OLHWDB2 and ICDAR-2013.

• Compared with state-of-the-art architecture, TCRN not only avoids the domain-specific knowledge for feature image extraction, but also attains higher training efficiency with a more compact model.

• Empirically, TCRN also outperforms the single recurrent neural network with faster prediction.

摘要

•We present a 3D touchless in-air handwritten Chinese text recognition system for the first time.•We constructed the first publicly available in-air handwritten Chinese text dataset, named IAHCT-UCAS2018, for the community.•A novel architecture, named the temporal convolutional recurrent network, is proposed for online handwritten Chinese text recognition, which achieves the competitive performance compared with the state-of-the-art methods on CASIA-OLHWDB2 and ICDAR-2013.•Compared with state-of-the-art architecture, TCRN not only avoids the domain-specific knowledge for feature image extraction, but also attains higher training efficiency with a more compact model.•Empirically, TCRN also outperforms the single recurrent neural network with faster prediction.

论文关键词：In-air handwriting,Handwritten chinese text recognition,Temporal convolutional recurrent networks

论文评审过程：Received 17 April 2019, Revised 1 August 2019, Accepted 26 August 2019, Available online 26 August 2019, Version of Record 2 September 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.107025