Preliminary Validation of a Cursive Handwriting Reconstruction Algorithm from a Sensorized Ink Pen.

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Educational Technology

Background:

• Handwriting difficulties in childhood can impact academic and social development.
• Objective quantification of handwriting performance is crucial for early intervention.
• Existing methods using Smart Ink Pens (SIP) capture the handwriting process but lack product detail.

Purpose of the Study:

• To develop and validate a trace reconstruction algorithm for handwriting using SIP signals.
• To assess the feasibility of reconstructing handwriting product from kinematic and force data.
• To evaluate the accuracy of reconstructed handwriting using Optical Character Recognition (OCR).

Main Methods:

• A trace reconstruction algorithm was developed using force and IMU signals from a SIP.
• 50 Italian 5th-grade students wrote 353 cursive words ('uno').
• Reconstructed traces were compared to actual scans using Google Vision OCR.

Main Results:

• OCR character recognition rates were 81.02% for scans and 59.49% for reconstructions.
• OCR character error rates were 21.48% for scans and 47.65% for reconstructions.
• 15% of reconstructions were misread in the opposite direction, indicating areas for algorithmic improvement.

Conclusions:

• The developed algorithm shows promise in reconstructing handwritten traces while preserving natural gesture.
• Preliminary results suggest the potential for objective handwriting assessment using ecological sensorized pens.
• Future improvements include increasing SIP sampling rate and refining trajectory reconstruction for better accuracy.