DSP techniques for narcotic detection using ion mobility spectrometry

Abstract

This paper evaluates the performance of several conventional Digital Signal Processing (DSP) techniques for peak detection in Ion Mobility Spectrometry (IMS). It deals with derivative, cross-correlation, curve fitting, and Hopfield neural networks methods. The performance and limitations of each method is discussed in terms of detection limit, probability of false alarm, and selectivity. The evaluation is done using real IMS field data obtained from injecting different amounts of Cocaine in the spectrometer. The paper also discusses a specific problem whereby a small peak may be present in the vicinity of a dominant one. It shows that conventional techniques are not capable of detecting peaks in these scenarios. A new DSP technique based on asymmetric adaptive curve fitting is proposed and its performance is evaluated using real field test data obtained from injecting various mixtures of Methamphetamine and Nicotine in the spectrometer.