[notepad]Principal Component Analysis for handwritten digit recognition

Loïc Abbès and I chose to work on handwritten character recognition for our term project. It was supervised by Laurence Likforman. Actually, we focused on the Most Significant Component Analysis, which preceeds the actual character classification.

Link Size Description
pdf icon 264 kB

Our report (in French)

January 2005 update: this is a reworked version of the initial report (same content, with a few corrections and improved presentation).

archive icon 8 kB

Compressed archive that gathers our whole work, including our programs (in Matlab), the generated images, our presentation (PowerPoint) and our report (along with its LaTeX sources).

January 2005 update: unfortunately the archive was lost! But I could recuperate the original report and reconstitute the scripts from its content. In the process, I converted the Matlab script in Octave scripts. This archive contains the Octave scripts and the LaTeX source of the report.

archive icon 8 MB

The digits used in this project. The data was published by NIST, more data, and their format, are available.

archive icon 28 kB

The labels corresponding to the digits used in this project.