Fig. 1. Original signals and reconstructed signals by LMD. (a1) Car cross original signal; (a2) car cross reconstructed signal;(b1) running original signal; (b2) running reconstructed signal; (c1) noise original signal; (c2) noise reconstructed signal; (d1) pickaxe original signal; (d2) pickaxe reconstructed signal; (e1) tapping original signal; (e2) tapping reconstructed signal

Fig. 2. Three-dimensional feature map of five different signals

Fig. 3. Flow chart of ensemble learning classification

Fig. 4. Implementation of AdaBoost

Fig. 5. Learning curves of decision tree and its AdaBoost classifier under different parameters. (a) Max_depth of decision tree; (b) number of base classifiers

Fig. 6. SVM learning curves. (a) Penalty coefficient C; (b) core parameter γ

Fig. 7. Precision, recall and F₁-score for 10-fold cross-validation with different classifiers

Fig. 8. Experimental flowchart

Fig. 9. Confusion matrixes of test samples. (a) SVM; (b) DTC; (c) AdaBoost-DTC

Fig. 10. Fiber optical identification true positive rates based on three different classifiers

Table 1. Important parameters and optimal parameter values of different classifiers