用户的技能水平对人机交互有着很大的影响，对于新手用户来说，学习新的界面布局是一项有挑战性的任务。我们通过随机化虚拟键盘的方式来引发用户在文本输入任务上的不同技能水平，包括静态随机键盘、动态随机键盘与传统Qwerty键盘三种布局，使用眼动设备、运动捕捉系统结合移动应用程序记录被试在文本输入过程中的触摸操作行为、眼动，与手指运动数据，进而分析被试在不同熟练度下的文本输入运动策略，包括新手阶段与学习过渡阶段的行为策略变化。研究发现，被试在新手阶段下视线主要集中在键盘区域进行按键的搜索与手指的定位，同时也会快速而频繁地校对输入结果，手指运动效率较低，然而被试在谨慎的文本输入策略下会产生较少的错误；被试在学习过渡阶段下视线也主要集中在键盘区域，以学习整体布局，文本输入策略由保守逐渐变得激进，错误率开始提高，并且由于校对工作较少且并不细致，因此会影响整体输入准确性；被试在熟练阶段下由于掌握了键盘布局的认知图式，定位按键所需的注意资源需求大大降低，按键搜索与手指运动路径引导更加直接而明确，因此注意力会相对更多地放在文本显示区域进行校对工作，虽然可能会出现大量的错误，但有赖于注意力资源向文本校对倾斜，修改错误的成本下降，整体的文本输入效率得到显著提升。对于移动设备的使用来说，准确高效的人机交互与信息交换是影响用户体验的重要因素，了解用户在使用不同布局的键盘时如何改变自己的文本输入行为与策略，有助于帮助研究人员和设计人员找到更高效的方法来改进文本输入界面，从而进一步有效地提高人机交互效率。研究获取的数据集和对应行为分析，也为接下来的文本输入手眼行为建模和未来进一步的自适应键盘设计工作提供理论和数据基础。

User's expertise has a great impact on human–computer interaction. Especially for novice users, learning a new interface layout is a challenging task. We randomized the layout of the Qwerty keyboard to observe the user behavior in different skill levels of the text input task, including statically randomized keyboard, dynamically randomized keyboard and Qwerty layout keyboard. Eye movement data, finger movement data and typing data were collected by eye trackers, motion capture systems and mobile apps, respectively. Then we analyzed the typing performance and eye–hand movements of the participants at different stages, including the change of behavior of novice stage and the transition stage. The results showed that the eyes of the participants in the novice stage mainly focused on the keyboard area for searching and locating, and they would proofread the input results quickly and frequently, the efficiency of finger movement was low. However, fewer errors were made when the subjects applied a cautious text input strategy. In the transition stage of learning, the participants also mainly focus on the keyboard area to learn the overall layout. The text input strategy gradually changed from conservative to aggressive, and the error rate began to increase. in addition, the accuracy of the overall input would be affected due to the less and not meticulous proofreading. Participants under the skilled stage due to the cognitive schema of the keyboard layout, attention resources for locating keys is greatly reduced, the path of eye to search and guide finger movement are more direct and explicit, participants will relatively pay more attention on the proofreading text area, depended on the allocation of attention resources to text proofreading, even if there are a lot of errors. The cost of correcting errors was effectively reduced and the overall efficiency is significantly improved. For the use of mobile devices, the accuracy and efficiency of human–computer interaction and information exchange is the important factor affect the user experience, understand the user how to change their behavior and strategy when they use different keyboard, it will help the researchers and designers to find more efficient ways to improve text input interface, Thus the human–computer interaction efficiency was improved. The dataset and behavior analysis obtained in this study also provide theoretical and data basis for the modeling of typing eye–hand behavior and further keyboard design in the future.