Abstract: To predict and improve human performance in temporary segmentation Chinese handwriting recognizer, static and dynamic performance models were set up. Experiment 1 verified and modified these two models, and an integrated model (D=N (WT+ST+R+439) + N[(1/RA)-1] [(0.816WT- 51.94 -439)+WT+ST+R+439] –(N/24)4973) was obtained. Experiment 2 further indicated that the integrated model could also account for the variance of the human performance data in another handwriting recognizer satisfactorily (R square>0.7). The application of the integrated model suggested that: (1) When the recognition accuracy (RA) and the number of characters inputted (N) were constants, the order of influence of the four factors (repairing time (T), average writing time for each character (WT), segmentation time (ST), recognition time (R)) was WT=ST=R>T; (2) When these four factors were constants, 1% improvement of recognition accuracy will save the task completion time at least 1000 ms

Key words: words mathematics psychological modeling, Chinese handwriting recognizer, human performance modeling, segmentation time