Abstract:

In order to study a rapid identification methods of feather types, the electronic nose was used in distinguishing the goose down and duck down, thus a qualitative feather types prediction model was established. The average response values (96s-98s) of the feather samples were extracted as characteristic walues, the principal component and linear discriminant method were used for qualitative judgement. The identification power was discussed between different pattern recognition methods. A down types prediction model was established by partial least squares. The results proved that the differentiation on the goose down and duck down is 89.2 % by principal component analysis, and the cumulative variance contribution ratio reaches 99.9 %. For the linear discriminant method, the first principal differentiated component contribution ratio is 90.63 %. For the partial least squares method, the classification success ratio is 97.5 % for the training sets and 90 % for the testing sets.

Key words: electronic nose, goose down, duck down, identification method, linear discriminantanalysis, partial least squares analysis