In this paper, a new method to enhance sensor selectivity is described. A flow modulation system driven by a PC-controlled peristaltic pump has been designed to feed a sensor chamber with different vapors. 45 measurements where performed comprising five different species (benzene, toluene, o-xylene, methanol and para-xylene) in three different concentrations (20, 200, 2000 ppm). Using frequency domain techniques and neural networks, the system was able to reach a 92% classification success rate when identifying all five vapors despite concentration was not constant and a single sensor was used. Moreover, when amplitude and variance information were removed from sensor transient signals, a 62% success rate was achieved, proving that the transient waveform has additional information that helps to enhance selectivity.