Using Electronic Nose in Forensic Odor Analysis

Forensic odor analysis often uses biodetectors (usually dogs) to perform odor matching, such as when determining whether two odor samples correspond to each other. Biodetectors can be hard to train, expensive to maintain and time-consuming in application. A task of finding a sample corresponding to a given one out of several samples is prohibitively expensive with large enough number of samples. An alternative to biodetector is using artificial odor analyzer, so called e-nose. We propose a two-step method for finding a corresponding sample out of several samples: first picking few candidates using e-nose then applying biodetector to find the required sample out of the already chosen candidates. Provided e-nose has high enough performance characteristics, this approach can make the task solvable in practice. We also calculate theoretical performance of this method as well as generalize the method into an abstract cascade classifier similar to the one used in Viola-Jones algorithm and calculate its theoretical performance.