An apparatus which transforms visual information to auditory output, was considered and designed. The system transforms visual image into 25×25 dot elements using a CCD image sensor and produces sound output. The vertical position of elements are displayed by frequency codes, and horizontal position of them are presented to each ear through binaural headphone by the ratio of sound amplitude.
Experiment showed that optimum sound presentation time was over 2.0 seconds for letter image recognition. Also the best case achieved 90 percent correct response in Katakana recognition, and 91 percent correct response in alphabet recognition. According to the results, similarity of letter images were analyzed by the hierarchical clustering scheme and discussed through the Dendrograms.
These notions will be applied for visual substitution system on a pattern recognition and an obstacle detection method for the blind.
View full abstract