A Bayesian Approach of Forensic Handwriting Evidence
Amy Crawford
A Bayesian Approach to Forensic
Handwriting Evidence
Amy Crawford
Center for Statistics and Applications in Forensic Evidence
Iowa State University
July 30, 2018
Statistics in Forensics
- 2016 PCAST Report on Forensic
- President’s Council of Advisors on Science and Technology recommended actions be taken to strengthen forensic science and called attention to the scientific validity of the analysis of some forms of forensic evidence.
- Questioned documents were an evidence form called to attention.
- Questioned document examiners (QDEs)
- Proficiency testing
- Expert Opinion
A Situation
- Classroom (10 students)
- Threat letter
- Task for QDEs
Computer Vision Lab (CVL)
- Vienna University of Technology
- 310 writers
- 7 handwritten texts
- 6 in English, 1 in German
Sample writing: 
Graphemes
Example of a grapheme in the FLASH ID interface.
SCREENSHOT
Grouping Graphemes
- (a) graphemes with 4 nodes and binary connection code 112 (label 4_112).
- (b) graphemes with 4 nodes and different binary connection codes.
- (c) a grapheme with 6 nodes.
Data
Grapheme Selection
- Graphemes are placed in order of importance for predictive analysis using random forest.
A few of the most “important” graphemes.
Analysis of Questioned Writing
How Many Graphemes?
Writing Samples
TOP: Writer #1
BOTTOM: Writer #6
How Many Graphemes?
Writer 3?
Writing Samples
TOP: Writer #3
BOTTOM: Writer #4
Density Estimates of Posterior Distn. for πw’s
The Situation
- Classroom (10 students)
- Threat letter
