|
Our Company presents SDK for application developers that require handwriting text
recognition. This SDK is based on recognition kernel "PenReaderLite". This kernel
allows to enter separate (not joint) handwritten multi-letter text.
Recognizing kernel "PenReaderLite" has unique features, you can't find elsewhere. The
main features are:
-
Multiple languages. At present kernel supports 28 languages, including
all European, Russian, Greek, Hebrew and graffity (in Palm OS standard). We
advisedly included graffiti to demonstrate universality of this technology that
allows us to add easily new graphic systems to our kernel. For example, if you
don't find a language you need among the list of supported languages, by your
request we'll on-the-fly add it to the kernel. Adding a new language slightly
increases computer resource demand.
-
Retraining. In contrast to the traditional system of recognition, where
retraining was the most difficult procedure and could be fulfilled only by
developers because of bad process control, we included retrain procedures into
our HWR SDK. For example, if you or your program product user writes letter 'A'
by his original way, not alike any of standard samples available in the base of
standard writing, he has to write this symbol only 9 times in his own manner in
order for the system to begin recognizing the way of writing unknown earlier.
-
Dynamic self-learning. If you want to increase quality of recognizing,
you can use the present in system the dynamic self-instruction mode. In this
mode system analyzes reliability of recognizing and in case of need displays
help menu, containing possible variants of replies. User only needs to confirm
the right version. Several such confirms are enough for system to manage a hard
or new writing.
-
Multi-platform system. At present HWR SDK exists for Windows 9x/2000/XP,
Windows CE/Pocket PC and Pocket PC 2002, Linux, Epoc 32/Symbian OS.
Now release SDK for Palm OS is in preparation. Taking into consideration our
great experience in different platform support, by your request we can port our
kernel with any of the 32-bit accessed platform, for which C++ compiler is
available.
-
Resource simple tastes. Recognizing kernel is provided in the form of
dynamic library. Depending on the processor it takes 140 Kb (SH3), 200 Kb
(MIPS), 160 Kb (ARM), 160 Kb (Intel x86). Besides library HWR SDK contains data
file that keeps current system condition. Depending on the number of the
supported languages it may take size from 600 Kb to 1200 Kb.
-
High capability. Thanks to the careful optimization our system shows high
results even with a low-powered processors that are usually used in pocket PC.
For example, nowadays the most popular is PPC Pocket PC with ARM SA1110 kernel
that achieves speed of recognizing 100 - 200 symbols / sec.
From our website you can download "PenTrainer" demo version for Pocket PC/Pocket PC
2002. It uses the same kernel that HWR SDK contains, but it is linked as a static
library and it is simply included to the demo-application. There is not another
difference between kernel used in "PenTrainer" and HWR SDK kernel except the way of
linking. With the help of "PenTrainer" you can appreciate:
- Capability of learning the individual way of writing. "PenTrainer" demonstrates
full cycle of retraining and different dynamic self-learning ways of writing.
- Full kit of SDK capabilities. All the functions, SDK has, are used in
"PenTrainer" for the demonstration a way of use kernel and capability of adapting.
Download PenTrainer (Pocket PC)
Download PenTrainer (Pocket PC 2002)
|
Главная
Новости в формате RSS
PenReader recognition software became a Finalist in the "Utilities - Handwriting" category in Smartphone & Pocket PC Magazine’s Sixth Annual Best Software Awards 2007
PenReader
выбран финалистом конкурса "Лучший продукт 2004 года" журналом Pocket PC Magazine.
PenReader
выбран финалистом конкурса "Лучший продукт 2003 года" журналом
Pocket PC Magazine.
PenReader
номинирован на награду "Лучший продукт" 2002 года журналом Pocket PC Magazine.
О компании