Writer Screening Handwriting Recognition
1. Objectives
Our aim is to develop and construct an interactive, automated writer screening station (WSS). The purpose of the WSS is the association of handwritten samples of an unknown writer with existing handwriting samples in a large database.
The input to the WSS is a handwritten sample (ink or graphite on paper). The output is:
- A computer-screen image of the input sample
- A list of samples in the WSS's database which are similar to the input sample.
- A computer-screen image of the most likely database document that matches the input sample, and the and the other database documents selected by the human operator.
- Portions of the computer-screen images of user-selected documents, magnified and shown side by side.
- Pertinent classification information, e.g., degree of similarity between documents, degree of confidence in classification decisions, peripheral information about the document's origin, date, and other relevant identification data.
In addition, the WSS will have the capability to obtain information and submit information to algorithms which perform document screening on the basis of context. The WSS algorithms are text-insensitive.
2. Specific Goals
The project has five concrete goals.
- to allow semi-automatic search in a large database for documents that are similar to a handwritten note, in interaction with a human operator; to identify and retrieve documents that are deemed similar to a questioned document;
- to display questioned documents, database documents and selected portions of these documents on large computer screens, and allow manipulations such as magnification, side-by-side comparison, x-y translation, etc.;
- to collect and display classification data and characterizing parameters on documents and portions of documents, such as: document and letter features (e.g. moments, correlation indices), degrees of similarity between documents, and level of confidence in classification decisions of the system;
- to incorporate and integrate into the system methods and techniques used by human document examiners; to use operational experience of document examiners in the classification algorithms and in the document database representation;
- to allow integrated operation of the WSS with algorithms and computer routines which screen documents according to text-sensitive criteria and peripheral document information (e.g. place where document was collected);
These objectives will be achieved through the development, synthesis and testing of a UNIX-based station.
This project is supported by the Department of Justice.
Members Working on Writer Screening
Main Software Development: Yang Wang
Additional Research: Tarek Qubain