Deeply embedded handwriting recognition
M. Hervieu, J.-Y. Brunel · 2002
The PHRASES system described generates handwriting recognizers to fit on one chip with the embedded processor of a personal digital assistant. It resorts to an artificial neural network symbol classifier that is backed up by a dictionary search at word level. It uses the chip-synthesis ALMA methodology to produce embeddable implementations which are tuned to application specific requirements such as the resolution of the input tablet or the dictionary size.>