THE METHOD OF BASELINE MANIPULATION TO OVERCOME THE SENSOR DRIFT ON GAS SENSOR TEST FOR HERBAL DRINKS DISCRIMINATION
Abstract Gas sensor system is widely used for the detection of aroma. The main problem in this system is the sensor drift that makes poor reproducibility of the sensor. The reproducibility of the sensor can be improved by applying the feature selection of the sensor’s output response and baseline manipulation. This research focused on determining methods that can reduce the dift sensor of gas sensor by using basaeline manipulation and selecting the optimal type of baseline manipulation when gas sensor system detects three different types of herbal drinks. The data that have been feature selected were then applied to three different types of baseline manipulation (differential, relative and fractional) and inserted into the pattern recognition system, Principal Component Analysis (PCA). From the analysis of PCA baseline manipulation that gives optimal results is differential one with the value of PC1 82.71%. This shows that differential baseline manipulation is effective in reducing the occurrence of sensor drift. Keywords: electronic nose, gas sensor, baseline manipulation, feature selection