Fig. 1. Decomposition tree of wavelet packet for three layers

Fig. 2. Experiment steps

Fig. 3. Time-domain signals of different-types events. (a) Reflection event; (b) non-reflection event

Fig. 4. Root mean square errors calculated with different basic wavelet in different scales. (a) Scale 3; (b) scale 4; (c) scale 5; (d) scale 6

Fig. 5. 64-dimensional normalized feature vectors of different-types event signals (main picture is normal, and inserted figure is magnified to 100 times). (a) Reflection event; (b) non-reflection event

Fig. 6. Test results of support vector machine and back propagation neural network for different numbers of training samples. (a) Accurate ratio; (b) convergence time