Neuro-electronic nose system under temperature and humidity controlled environment

An electronic nose that has been successfully applied to classify various smokes is developed by adding a, mechanism to reduce the effect from temperature and humidity of the air. By controlling these factors, the signals from the same smell in every repetition are highly correlated. Thus, only a single data from each kind of smell that has the highest similarity index (SI) value to its repetition data is selected as a training data for the error back propagation 0 neural network (BPNN). A method called slope maximum mean (SMM) is applied to reduce the dimension of experimental data before being analyzed by the BPNN and 99.8% of the tested data are correctly classified.