Air-writing through hand or fingertip is a functional and attractive mechanism. Since there is usually no pen-up and pen-down in air-writing, trajectory of numbers and words in the air-writing will be a connected series of characters (ligature Stroke). Identification of legitimate characters such as digits or letters inside a ligature Stroke is one of the most important challenges faced in this area. By solving these challenges, there will be more uses in future. In this work, the color and depth images of the Kinect sensor are used to identify the user’s air-writing, which includes the digits and numbers of the Persian language. To extract a feature vector from the trajectory, we propose a simple but very effective method, called slope variations detection, which is robust to variations of size, translation, and rotation of the trajectory. Also, a novel analytical classifier is proposed to map a vector to a character. This classifier has higher speed and accuracy than traditional classifiers, such as SVM, HMM, and K Nearest Neighbors. Experimental results show that the average recognition rate for digits and numbers of Persian language is 98% which is quite acceptable for a practical system.

中文翻译：

---

一种新型分类器的波斯数字空写识别系统

通过手或指尖进行空中书写是一种实用且有吸引力的机制。由于空中书写通常没有上下笔，因此空中书写中数字和单词的轨迹将是一个连接的字符序列（连字笔画）。在连字笔画中识别合法字符（例如数字或字母）是该领域面临的最重要挑战之一。通过解决这些挑战，未来会有更多用途。在这项工作中，Kinect 传感器的颜色和深度图像用于识别用户的空中书写，其中包括波斯语的数字和数字。为了从轨迹中提取特征向量，我们提出了一种简单但非常有效的方法，称为斜率变化检测，它对轨迹的大小、平移和旋转的变化具有鲁棒性。还，提出了一种新的分析分类器来将向量映射到字符。该分类器比传统分类器（如 SVM、HMM 和 K 最近邻）具有更高的速度和准确度。实验结果表明，波斯语数字和数字的平均识别率为98%，在实际系统中是可以接受的。