Purchase this article with an account.
K. Yang, X. Chai, J. Fan, Q. Ren; Recognition of Chinese Characters With a Limited Number of Pixels Using Simulated Prosthestic Vision Based on Complexity Analysis. Invest. Ophthalmol. Vis. Sci. 2008;49(13):3039. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
It is of great importance to investigate the minimum resolution for recognition of pixelized Chinese characters in the research of visual prosthesis. In this study, the influences of pixel number and complexity on recognition of Chinese characters and the minimum requirement on the reading of pixelized Chinese characters were investigated.
The experimental set-up included a head-mounted display (HMD) and a Dell computer with a self-developed program. We calculated the complexity of 631 most commonly used Chinese characters by black pixel algorithm. The pixelized images of Chinese Character with different resolution (6*6, 8*8, 10*10, and 12*12) were rendered on the HMD one by one against black background. Forty volunteers from 26 to 48 years old were recruited. The testing sequence of the pixel arrays was 6*6, 8*8, 10*10, and 12*12. The subjects were requested to pronounce the character if it was legible, and say "no" if it was illegible. Recognition accuracy of each character was recorded.
For the characters with 6*6 pixels, the recognition accuracy decreased quickly from 77.5±6.61% to 17.13±11.93% with the increment of complexity from the range of 0.00-0.16 to 0.20-0.24. When the complexity was higher than 0.32, the Chinese character became almost illegible with the recognition accuracy dropped to 1.13±2.37%. For the characters with 8*8 pixels, the recognition accuracy declined from 96.63±5.41% to 74.25±12.88%, 49±12.95%, and 25.88±11.72%, as the complexity increased from the range of 0.00-0.16 to 0.24-0.28, 0.28-0.32 and 0.32-0.36, respectively. The recognition accuracy was 94.88±3.44% for 10*10 pixels and 99.25±2.11% for 12*12 pixels at complexity less than 0.28, and reached 77.13±10.24% for 10*10 pixels and 93.88±5.30% for 12*12 pixels at complexity higher than 0.26. The recognition accuracy of high resolution characters was higher than that of low resolution characters, while the recognition accuracy decreased with the increment of complexity.
The results indicate that both pixel number and complexity of Chinese characters have significant impacts on the recognition of Chinese characters. 10*10 and 12*12 pixels array are sufficient for recognition of Chinese characters for a visual prosthesis.
This PDF is available to Subscribers Only