An experimental performance evaluation of a novel radio-transmitter identification system under diverse environmental conditions

In this paper, the effects of environmental conditions on a radio-transmitter identification system are investigated. Transmissions from 10 commercial VHF FM transmitters operating under varying environmental conditions are analyzed using an experimental setup. It is observed that a change in environmental conditions such as battery voltage or ambient temperature causes the feature vectors to spread over the feature space, resulting in classification performance degradation. It is experimentally shown that performance degradation resulting from changing environmental conditions can be compensated for by training the system with data collected in a range of battery-voltage and ambient temperature levels. The signal-to-noise ratio versus probability of correct classification is also determined for systems operating in noisy channels in order to obtain a confidence measure for the classification decision to be made. The transient features are extracted using a technique based on the amplitudes and phase changes of the received signals. This technique is well suited to radio-transmitter turn-on transient signals, and classification with these features, as shown, outperforms current techniques.