A radical-partitioned neural network system using a modified sigmoid function and a weight-dotted radical selector for large-volume Chinese character recognition VLSI

This paper presents a radical-partitioned neural network system using a modified sigmoid function and a weight-dotted radical selector for large-volume Chinese characters recognition VLSI. With a modified sigmoid function and the weight-dotted radical selector, the recognition rate of 1000 radical-partitioned Chinese characters can be enhanced to 90% from 70% for the input samples with 15% random errors as compared to the system without it.<
>