Institut de Recherche Criminelle de la Gendarmerie Nationale, Caserne Lange, 5 boulevard de l'Hautil, BP 20312 Pontoise, 95037 Cergy, Pontoise CEDEX, France; UMR CBI 8231, CNRS, Laboratoire Sciences Analytiques Bioanalytiques et Miniaturisation, ESPCI Paris, PSL Research University, 10 rue Vauquelin, Paris, France.
Equipe de Statistique Appliquée, ESPCI Paris, PSL Research University, INSERM, UMRS 1158, Neurophysiologie Respiratoire Expérimentale et Clinique, 10 rue Vauquelin, Paris, France.
J Chromatogr B Analyt Technol Biomed Life Sci. 2018 Aug 15;1092:379-385. doi: 10.1016/j.jchromb.2018.06.018. Epub 2018 Jun 15.
A new method for identifying people by their odor is proposed. In this approach, subjects are characterized by a GC × GC-MS chromatogram of a sample of their hand odor. The method is based on the definition of a distance between odor chromatograms and the application of Bayesian hypothesis testing. Using a calibration panel of subjects for whom several odor chromatograms are available, the densities of the distance between chromatograms of the same person, and between chromatograms of different persons are estimated. Given the distance between a reference and a query chromatogram, the Bayesian framework provides an estimate of the probability that the corresponding two odor samples come from the same person. We tested the method on a panel that is fully independent from the calibration panel, with promising results for forensic applications.
提出了一种通过气味识别个体的新方法。在该方法中，通过对受试者手部气味样本的 GC×GC-MS 色谱图进行特征描述。该方法基于定义气味色谱图之间的距离，并应用贝叶斯假设检验。使用校准面板中的受试者，这些受试者有多个气味色谱图可用，估计同一人的色谱图之间以及不同人的色谱图之间距离的密度。给定参考色谱图和查询色谱图之间的距离，贝叶斯框架提供了相应两个气味样本来自同一个人的概率估计。我们在一个与校准面板完全独立的面板上测试了该方法，对于法医学应用，结果很有前景。
