1. Pushing Out the Frontiers of Forensic Science
Application of Chemometrics and Advanced Pattern Recognition to Trace Evidence Analysis
Pushing Out the Frontiers of Forensic Science

2. Outline Admissibility of Scientific Evidence is a problem!
Frye and the Daubert Standards
How chemistry, math and computers can help forensic science
Current Projects At John Jay:
Gasoline
Tool Marks
Footwear
Questioned Documents

3. Admissibility of scientific evidence!
Principal legal standards: Frye and Daubert
Frye (1923) – Testimony offered as “scientific” must “...have gained general acceptance in the particular field in which it belongs”.
North Carolina and New York are still a “Frye States”

4. Frye and Daubert
Daubert (1993)- Judges are the “gatekeepers” of scientific evidence.
Must determine if the science is reliable
Has empirical testing been done?
Falsifiability
Has the science been subject to peer review?
Are there known error rates?
Is there general acceptance?
Federal Government and 26(-ish) States are “Daubert States”

5. Raising Standards with Data and Statistics
DNA profiling the most successful application of statistics in forensic science.
Responsible for current interest in “raising standards” of other branches in forensics.
No protocols for the application of statistics to physical evidence.
Our goal: application of objective, numerical computational pattern comparison to physical evidence

6. What Statistics Can Be Used?
Statistical pattern comparison!
Modern algorithms are called machine learning
Idea is to measure features of the physical evidence that characterize it
Train algorithm to recognize “major” differences between groups of features
while taking into account
natural variation and measurement error.

7. Statistical Methods Principal Component Analysis (PCA)
Method minimizes the dimensionality of the data by discounting variables with minimal contributions to the overall spread of the data.
3D PCA
2D PCA
Raw Data

8. Canonical Variate Analysis (CVA)
Maximize the difference between groups by exploiting the inter and intra group variance in determining clustering.
2D PCA
2D CVA
Raw Data

9. For Classifying Data: Neural Networks Support Vector Machines
“Train” the decision rules:
Support Vector Machines
“Train” the decision rules:
0 ≤ li ≤ C i = {1, …, n}

10. Error Rates: Hold-Out Bootstrap
Use a big test set with trained algorithm
Hold-one-out most common
Hold-none-out generates apparent error rate. Biased and overly optimistic
Bootstrap
Sample from data set many times (n-bootstrapped samples)
Train algorithm n times with n-bootstrapped samples
Use each set of bootstrap trained decision rules on entire data set
Average the n-error rates together - bootstrap error rate

11. Conformal Prediction Theory
New, but has roots in 1960’s with Kolmogorov’s ideas on randomness and algorithmic complexity.
Can be used with any statistical pattern classification algorithm.
Independent of data’s underlying probability distribution.
This is a very important property for forensic tool mark analysis!!
For identification of patterns, method produces
Level of confidence, 1-ε
Measure of how likely identification is to be correct
Results are valid:

12. Fire Debris Analysis Casework
Liquid gasoline samples recovered during investigation:
Unknown history
Subjected to various real world conditions.
If an individual sample can be discriminated from the larger group, this can be of forensic interest.
Gas-Chromatography Commonly Used to ID gas.
Peak comparisons of chromatograms difficult and time consuming.
Does “eye-balling” satisfy Daubert, or even Frye .....????

13. Study Design
This study was undertaken to examine the variability of gasoline components in
Twenty liquid gasoline samples
Samples from fire investigations in the New York City area
All samples analyzed using Gas Chromatography-Mass Spectrometry
Keto and Wineman target compounds
Fifteen peaks were chosen in this study that represented the common components present in gasoline.
Normalized GC-MS peak areas were utilized to test the discrimination potential of multiple multivariate methods for discrimination.

14. Chosen Peaks

15. 2D PCA 97.3% variance retained
Avg. LDA HOO correct classification rate: 83%

16. 3D PCA 98.7% variance retained
Avg. LDA HOO correct classification rate: 88%

17. 2D CVA
Avg. LDA HOO correct classification rate: 92%

18. 3D CVA
Avg. correct classification rate: 100%.

19. G. Petillo 5/8” Consecutively manufactured chisels
Known Match Comparisons
G. Petillo

20. Current Approach For Striated Tool Marks
Obtain striation pattern profiles form 3D confocal microscopy

21. Glock 19 firing pin impression
Primer shear
Glock 19 firing pin impression

22. 3D confocal image of entire shear pattern

23. Shear marks on primer of two different Glock 19s

24. Mean “waviness” profile:
Mean total profile:
Mean “waviness” profile:
Mean “roughness” profile:

25. 3D PCA-SVM Bootstrap error rate ~1%:

26. Accidental Patterns on Footwear
Shoe prints contain marks and patterns due to various circumstances that can be used to distinguish one shoe print from another.
How reliable are the accidental patterns for identifying particular shoes?

27. Facial Recognition Approach to Accidental Pattern Identification
3D PCA
59.7% of variance

28. Questioned Documents: Photocopier Identification
Mordente, Gestring, Tytell
Photocopy of a
blank sheet of paper

29. Just Getting Started: Things to Come
Dust
Soil
Wrenches
Chisels
Hammers
Tire Tracks
Hair
Blood Spatter
Gun Shot
Residue

30. Acknowledgements National Institute of Justice
New York City Police Department Crime Lab
John Jay College of Criminal Justice
Research Team:
Mr. Peter Diaczuk
Ms. Carol Gambino
Dr. James Hamby
Dr. Thomas Kubic
Off. Patrick McLaughlin
Mr. Jerry Petillo
Mr. Nicholas Petraco
Dr. Peter A. Pizzola
Dr. Graham Rankin
Dr. Jacqueline Speir
Dr. Peter Shenkin
Mr. Peter Tytell
Helen Chan
Manny Chaparro
Aurora Ghita
Eric Gosslin
Frani Kammerman
Brooke Kammrath
Loretta Kuo
Dale Purcel
Stephanie Pollut
Rebecca Smith
Elizabeth Willie
Chris Singh
Melodie Yu