New pattern recognition techniques for searching infrared (IR) spectral libraries of the Paint Data Query (PDQ) automotive paint database will be discussed.  Applying a wavelet packet tree to denoise and deconvolve FTIR spectra of clear coats by decomposing each spectrum into wavelet coefficients which represent the sample's constituent frequency, a genetic algorithm (GA) for feature selection and pattern recognition will be used to identify wavelet coefficients characteristic of the model and manufacturer of the automobile from which a clear coat paint sample has been obtained.  All too often, only a clear coat paint smear is left at the crime scene.  In these cases, the text based portion of the PDQ database cannot be used to identify the motor vehicle.  Even in challenging trials where the samples evaluated were all the same manufacturer (Chrysler) with a limited production year range, the respective model and manufacturing plants could be correctly identified.  Search prefilters for infrared spectra, which are the focus of this presentation, are necessary to extract investigative lead information from a clear coat paint smear.  Information derived from the proposed pattern recognition searches will serve to quantify the general discrimination power of original automotive paint comparisons encountered in casework, and will further efforts to succinctly communicate the significance of the evidence to the courts. Addressing these concerns is a direct response to Recommendation 3 of the National Academies' February 2009 report, �Strengthening Forensic Science in the United States: A Path Forward.�.  The potential to perform similarity searching using these search prefilters and a library search algorithm rotted on the cross correlation function will also be discussed.

Wavelets and genetic algorithms for pattern recognition based library searching of the PDQ database

242nd National Meeting (2011)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.