Comparison of Difference, Relative and Fractional Methods for Classification of The Black Tea Based on Electronic Nose

Abstract

The ability of electronic nose (e-nose) in classifying is determined by methods used in preprocessing, features extraction, and pattern recognition. Each method has advantages in choosing unique features that are hidden in sensor response. Comparison of the methods is used to obtain the best approach in preprocessing. The aroma of black teas (Broken Orange Pekoe, Broken Pokoe II, and Bohea) was measured 160 times. Sensor response is processed with three preprocessing models, and features are extracted using the maximum method. The best method is determined based on the classification of three black teas that are formed, and it was carried out after data clustering was successfully made with principal component analysis (PCA). As a result, three black teas can be clustered with 98.0% of total variant of data. In general, classification can be done with these methods. However, the best classification uses difference because signal amplitude high, difference amplitude between signals and noise are small.