Datasets
Standard Dataset
K-AirWrite: Real-Time Continuous Air Handwriting Recognition Using Kinect
- Citation Author(s):
- Submitted by:
- zhixiong Yang
- Last updated:
- Mon, 04/15/2024 - 10:35
- DOI:
- 10.21227/yrfr-sg43
- License:
- Categories:
- Keywords:
Abstract
This dataset contains continuous gesture data for both Chinese and English, including 14 Chinese characters and 4 English words. The Chinese characters are: 不 (bù), 程 (chéng), 刀 (dāo), 工 (gōng), 古 (gǔ), 今 (jīn), 力 (lì), 刘 (liú), 木 (mù), 石 (shí), 土 (tǔ), 外 (wài), 中 (zhōng), 乙 (yǐ). The English words included are: 'can', 'NO', 'Who', 'yes'.
Using a Kinect device, 10 participants (5 men and 5 women) collected continuous gesture data for 14 Chinese characters and 4 English words in three different environments. The characters and words include: “不”, “程”, “刀”, “工”, “古”, “今”, “力”, “刘”, “木”, “石”, “土”, “外”, “中”, “乙”, “can”, “NO”, “Who”, “yes”. This data was processed using the M-YOLOv8 algorithm. After processing, the dataset contains 28,000 gestures for 'yes', with 70% designated as the training set and 30% as the test set