Fig 1.
Cole plot and frequency bands definition.
Fig 2.
BIS measurements (solid red line) containing the 6 identified types of measurement errors plotted in the impedance plane. Estimations of the clean measurements are represented by dashed blue lines.
Type-A error (A), Type-B error (B), Type-C error (C), Type-D error (D), Type-E error (E), Type-F error (F).
Table 1.
Relationship between spectral immittance components.
Fig 3.
Immittance components computation.
Fig 4.
Steps to calculate the true and fitted spectral immittance components.
Table 2.
Proposed features for BIS measurement classification.
Table 3.
Classes definition: all-at-once approach.
Table 4.
Classes definition: divide and conquer approach.
Fig 5.
Distribution of the BIS experimental measurements included in the database.
Distribution per application (A), distribution per measurement type (B).
Fig 6.
Mean and standard deviation (5 folds) of the classification error (%) obtained by the feature selection algorithm for different values of NFEAT in the all-at-once approach (7 classes).
Table 5.
Selected features with NFEAT = 7 in the two classification approaches.
Table 6.
Confusion matrix in all-at-once approach.
Classification error of 5.7%.
Fig 7.
Mean and standard deviation (5 folds) of the classification error (%) obtained by the feature selection algorithm for different values of NFEAT in the divide and conquer approach (5 classes).
Table 7.
Confusion matrix in divide and conquer approach, step 1.
Classification error of 0.17%.
Table 8.
Confusion matrix in divide and conquer approach, step 2 (BC).
Classification error of 0.93%.
Fig 8.
Feature f4 for samples of Type-B error and Type-C error in (A) and feature f8 for samples of Type-E error and Type-F error in (B). Samples are taken from different BIS applications.
Table 9.
Confusion matrix in divide and conquer approach, step 2 (EF).
Classification error of 0%.
Table 10.
Classification error of the two proposed approaches with NFEAT = 7.