PSYCH OpenIR  > 社会与工程心理学研究室
Thesis Advisor张侃
Degree Grantor中国科学院研究生院
Place of Conferral北京
Degree Discipline应用心理学
Keyword手写输入系统 用户绩效 绩效建模 识别正确率 分割时间

   用户绩效是手写输入系统成功的关键。找到影响手写输入系统的用户绩效的各种因素,并对这些因素进行数学与统计建模,是估计与提高用户绩效的基础。本文推导了基于时间分割(直接重写立即修改方式)与空间分割(选择候选项延迟修改方式)的汉字手写输入系统的用户绩效模型(见第二章(C 1-12)式与第三章(2-25)式),并用实验验证了这两个模型在实际商用软件与不同训练水平用户中的有效性。
为使用户绩效达到每书写24个汉字耗时至少减小200ms的目的,对于使用手写框进行输入的手写系统,在屏幕空间允许的前提下,建议采用4}8框输入为宜。在屏幕空间仅能容纳2--3框f l-寸,建议采用双框设计;对于全屏幕手写的输入系统,建议悔行书写字在7-8个左右;
当单位候选字与识别错误频率的关系满足倒数模型时:当用户的手写速度在1 OOOms/字以下时,采用候选字数为5个能有效地提高手写绩效;当用户的手写速度在l OOOms-2000ms/字时,采用候选字数为6个能有效地提高手写绩效;当用户的手写速度在2000ms/字以上时,采用候选字数为7个能有效地提高手写绩效;
提高系统的首选正确率相对提高二选正确率更能有效地提高用户绩效;Human performance is the key of the success of handwriting recognizes Finding out the exact factors which will determine the perfor;nance and set up quantitative models to describe their relationship with the performance, are the base of predict and improve the human performance in these systems. In this study, human performance model (see model (1一12)) in temporal-segmentation handwriting recognizes, and human performance model (,see model (2-25)) in spatial-segmentation handwriting recognizes were set up by statistic method and mathematic inference, Their validity in real commercial product and user with different training experiences were verified by the experiments in this study. Six modules based on that these two models could be combined flexiblely to predict the human performance in various kinds of handwriting recognizers. Moreover, some quantitative guidelines to the design of Chinese handwriting recognizes were suggested by the result of experiment or deduced from the mathematic inference of the two performance models in this study:

1) To insure the human performance was not impaired significantly in the temporal-segmentation recognizes, the segmentation time should be not longer than 1s; And the preferred range of the segmentation time in these systems is between 0.2s to 1s;
2) Assuming that the reduction of task completion time should be longer than 200ms in the process of inputting every 24 characters if improvement of human performance is efficient:
    .To the spatial-segmentation recognizes whose input space is limited, 4-8 windows for inputting characters is preferred in the design of the interface; When the space is limited to 2-3 windows in user interface, the design of double windows is better that the one with three windows;
.To the spatial-segmentation recognizes whose input space is the whole screen, it was suggested that 7-8 characters per line can improve the human performance efficiently;
3) If there is an inverse model between the possibility of 1 unit's reduction of error frequency and the l unit's increase of alternatives, when user's writing speed is higher than 1000ms per character, 5 alternatives to be chosen in the choice area are preferred; When the speed is between 1000ms to 2000ms per character, 6-alternative choice area is preferred; So does the 7-alternative choice area if the speed is lower than 2000ms per character;
4) The minimal efficient improvement of first-choice recognition lccuracy should be around 0.23%-0.27%. if the second-choice recognition accuracy is set as a constant; And the higher of the second-choice recognition accuracy, the bigger of minimal efficient improvement of first-choice recognition accuracy;.
5) The minimal efficient improvement of second-choice recognition accuracy should be around 0.30%-0.36%, if tlue first-choice recognition accuracy is set as a constant; And the higher of the first-choice recognition accuracy, the bigger of minimal efficient improvement of second-choice recognition  accuracy; The improvement of first-choice recognition accuracy will improve user performance more efficiently than the improvement of'second-choice recognition accuracy;

Document Type学位论文
Recommended Citation
GB/T 7714
吴昌旭. 手写汉字输入系挽用户绩效模型的推导、验证与应用[D]. 北京. 中国科学院研究生院,2002.
Files in This Item:
File Name/Size DocType Version Access License
吴昌旭.pdf(5210KB)学位论文 限制开放CC BY-NC-SAView Application Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[吴昌旭]'s Articles
Baidu academic
Similar articles in Baidu academic
[吴昌旭]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[吴昌旭]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 吴昌旭.pdf
Format: Adobe PDF
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.