Towards a Likelihood Ratio Approach for Bloodstain Pattern Analysis

13 Pages Posted: 3 Sep 2022

See all articles by Tong Zou

Tong Zou

University of California, Irvine

Hal S. Stern

University of California, Irvine - Department of Statistics

Abstract

In this work, we explore the application of likelihood ratio as a forensic evidence assessment tool to evaluate the causal mechanism of a bloodstain pattern. It is assumed that there are two competing hypotheses regarding the cause of a bloodstain pattern. The bloodstain patterns are represented as a collection of ellipses with each ellipses characterized by its location, size and orientation. Quantitative measures and features are derived to summarize key aspects of the patterns. A bivariate Gaussian model is chosen to estimate the distribution of features under a given hypothesis and thus approximate the likelihood of a pattern. Published data with 59 impact patterns and 55 gunshot patterns is used to train and evaluate the model. Results demonstrate the feasibility of the likelihood ratio approach for bloodstain pattern analysis. The results also hint at some of the challenges that need to be addressed for future use of the likelihood ratio approach for bloodstain pattern analysis.

Keywords: Image Processing, Feature Extraction, Statistical Modeling, Forensic Statistics

Suggested Citation

Zou, Tong and Stern, Hal S., Towards a Likelihood Ratio Approach for Bloodstain Pattern Analysis. Available at SSRN: https://ssrn.com/abstract=4209110 or http://dx.doi.org/10.2139/ssrn.4209110

Tong Zou (Contact Author)

University of California, Irvine ( email )

Division of Nephrology, University of California I
101 City Drive South, City Tower, Suite 400-ZOT;40
Orange, CA California 92868-3217
United States

Hal S. Stern

University of California, Irvine - Department of Statistics ( email )

Campus Drive
Irvine, CA California 62697-3125
United States

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Downloads
27
Abstract Views
249
PlumX Metrics