Recognition System 

of Japanese Characters

<< BACK  NEXT >>

System Description

+ Hardware

• Polhemus Sensor

• PC (OS: Windows 2000).

+ Software

• Visual C++ 6.0

• Matlab 5.3

• Simulink 4.0

• Neuronal Network Toolbox.

Polhemus Sensor

    The polhemus sensor provides 6 DOF (3-position and 3-orientation). This sensor consist of three elements:

1. Transimisor: This black box produces a magnetic field, which principal characteristic is to have a black tip sourface (direction x+ by default configuration).

2. Control Unit: This unit controls the transmisor and the sensor to provide, by a RS-232 connector attached to the computer,  the position and orientation of the sensor.

3. Sensor: This was attached to a pen. This sensor provides the information necessary to be used by a handwritting recognition system.

Simulink Diagrams

    The handwritting recognition system was developed using Matlab and Simulink. To test this system we defined 10 Japanese characters as shown below:

    The main block diagram is presented below. The selection constant is used select the desired output from the likelihood list (10 has the biggest probability output).

    The next simulink diagram presents the 3 main block of this system:

• Fastrack: adquire the information from the Polhemus Sensor.

• Quantization: transforms continous data into discrete data.

• HMM_evaluation: to evaluate on-line the sequence data. 

    The fastrack module, using the information from Polhemus sensor, computes the contact point position.

    This contact position is computed using the following rotation matrix:

    The quantization module, using a Self-Organization Map, transforms the continous data to discrete data with a sample frecuency of 30Hz. 

Last update 01.15.2004

These pages are maintained by Jorge Solis