The Investigation of Human Scent as a Biometric through Targeted and Untargeted Chemometric Analyses

Authors

Vidia A. Gokool, Florida International UniversityFollow

Document Type

Dissertation

Degree

Doctor of Philosophy (PhD)

Major/Program

Chemistry

First Advisor's Name

Kenneth G Furton

First Advisor's Committee Title

Committee Chair

Second Advisor's Name

Jose Almirall

Second Advisor's Committee Title

Committee Member

Third Advisor's Name

Lauryn DeGreeff

Third Advisor's Committee Title

Committee Member

Fourth Advisor's Name

Tan Li

Fourth Advisor's Committee Title

Committee Member

Fifth Advisor's Name

Xiaotang Wang

Fifth Advisor's Committee Title

Committee Member

Keywords

Human scent, biometric, forensic identifier, hand odor, chemometrics, trace vapor analysis, volatile organic compounds, VOC, HS-SPME, GC-MS

Date of Defense

10-25-2022

Abstract

An investigation into the association of human scent using laboratory instrumentation was conducted to establish foundational knowledge of human scent variation over time and determine the potential probative use of human scent as a biometric. This research was conducted using headspace-solid phase microextraction-gas chromatography-mass spectrometry (HS-SPME-GC-MS) as the instrumental means of analysis. In total, fifty-seven (57) participants partook in the study, donating their hand odor at varying time points and frequencies. The collected hand odor samples were analyzed by HS-SPME-GC-MS and interpreted using novel software programs and the application of chemometric and machine learning concepts.

This research resulted in the development of two-dimensional and three-dimensional approaches to interpreting and associating human hand odor samples. The investigation of intra-subject human scent variations revealed underlying commonalities between same-source samples (up to 35 days in age), noting that these human hand odor profiles become increasingly dissimilar to one another as more time elapses between sample collections. This time-based variation was seen to occur with all eight (8) donors who participated in this study; however, the rate of odor variation was not consistent between individuals.

The investigation of inter-subject variation and the pursuit of human scent association revealed multiple modalities of associating human scent profiles. These measures provided moderate associative abilities on their own; however, when combined, the joint use of these approaches demonstrated increased discriminatory power.

Four optimized models were combined in a decision tree approach to perform human scent association. Their performance was determined using the 14-day intra-subject sample timepoint (intra-subject n=394, inter-subject n=13874). The decision tree utilized (1) the predictive outcome of a sex predictive linear discriminant analysis model, then (2) the measure of Spearman’s rank correlation test (SRC) applied to untargeted, reoccurring features of interest. Next, two tests were concurrently considered, (3) SRC applied to quantitated targeted analyte and (4) three-dimensional covariance calculation similarity measures. The final, combined, model scheme operates with a sensitivity/ true positive rate= 79.4%, specificity/ true negative rate = 97.6%, and false positive rate (FPR) = 7.0%. This work demonstrates the first published account of a forensic biometric scheme for the association of human odor samples.

Identifier

FIDC010860

ORCID

0000-0003-1867-1761

Recommended Citation

Gokool, Vidia A., "The Investigation of Human Scent as a Biometric through Targeted and Untargeted Chemometric Analyses" (2022). FIU Electronic Theses and Dissertations. 5218.
https://digitalcommons.fiu.edu/etd/5218