Introduction

MV-CISL (Multi-View Chinese Isolated Sign Language Dataset) is a large-scale multimodal dataset specialized in isolated Chinese Sign Language (CSL) recognition and lexical analysis. It provides high-quality video recordings of 60 unique CSL lexical gestures, paired with precise gloss-level annotations (individual sign unit labels) and corresponding Chinese character translations. The dataset captures each sign from three orthogonal camera views (frontal, left 45°, right 45°), with the frontal view additionally including depth maps and skeletal joint data for comprehensive spatial feature extraction.

The dataset focuses on daily-life lexical categories essential for basic communication. Each gesture is performed by 20 native CSL signers, incorporating natural variations in handshape and movement. As one of the few multi-view datasets specialized in isolated CSL units, MV-CISL addresses the critical need for standardized resources in lexical-level sign language processing. It supports fundamental research in isolated sign recognition (ISLR) and cross-view feature generalization, serving as key building blocks for developing assistive technologies and educational tools for the deaf community.

BP-CCSL (Business Processing Chinese Continuous Sign Language Dataset) is a Chinese continuous sign language dataset created by our team focusing on the business processing domain. This dataset emphasizes sign language expressions specific to business processing, providing targeted and practical training data to advance CSLR, particularly in domain-specific applications. To ensure the dataset's representativeness and the accuracy of sign language gestures, we collaborated with a school for the hearing impaired in Shanxi Province, China. Twenty students with hearing impairments (11 male and 9 female) participated under the guidance of professional sign language instructors. BP-CCSL consists of 2,000 sign language videos, including 1,600 samples from 16 subjects designated for training and 400 samples from 4 subjects allocated for the test set, maintaining a 4:1 training-to-test ratio. The dataset includes 33 Chinese words and 20 sentences, each consisting of 3 to 5 common words, with every sentence repeated five times by all 20 participants.

Download

The MV-CISL and BP-CCSL databases are released to universities and research institutes for research purpose only. To request the access right to the data resources, please follow the instructions below:

1.Download the MV-CISL Dataset Release Agreement or BP-CCSL Dataset Release Agreement;

2.Read all items and conditions carefully;

3.Complete it appropriately. Note that the agreement should be signed by a full-time staff member (that is, the student is not acceptable).

4.Please scan the signed agreement, send it to the E-mail (xuepeiyun@tyut.edu.cn). If you are a student, please also CC to the full-time staff member who sign the agreement.

Contact

If you have any questions about the dataset and our papers, please feel free to contact us:

Peiyun Xue, associate professor, TYUT, xuepeiyun@tyut.edu.cn

Click to download the application form.

1.application form-MV CISL

2.application form-BP CCSL