Evolutionary Metric-Learning-Based Recognition Algorithm for Online Isolated Persian/Arabic Characters, Reconstructed...

Area of Science:

• Human-Computer Interaction
• Sensor Technology
• Machine Learning

Background:

• Microelectromechanical systems (MEMS) technology enables advanced human-computer interaction tools like inertial pens.
• Inertial pens offer an inexpensive, hardware-independent alternative for writing input.
• Existing character recognition methods often rely on low-level inertial signal features.

Purpose of the Study:

• To develop a novel Persian/Arabic handwriting character recognition system for inertial-sensor-equipped pens.
• To leverage high-level geometrical features extracted from reconstructed motion trajectories.
• To improve character classification accuracy using an advanced metric learning technique.

Main Methods:

• Reconstructing inertial pen motion trajectories using inertial navigation system principles to estimate position signals.
• Extracting high-level geometrical features from the estimated position signals.
• Employing a genetic programming algorithm to calculate characteristic functions for each character, forming a metric kernel for classification.

Main Results:

• The proposed system successfully reconstructs pen motion and extracts relevant geometrical features.
• A novel metric learning approach significantly enhances character classification accuracy.
• Experimental results demonstrate superior performance compared to existing state-of-the-art methods for Persian/Arabic handwriting recognition.

Conclusions:

• The developed system offers an effective solution for recognizing Persian/Arabic handwriting using inertial pens.
• The integration of inertial navigation principles and metric learning advances handwriting recognition technology.
• This approach provides a robust and accurate method for character classification in human-computer interaction.