APPLICATIONS OF DATA MINING TECHNIQUES TO AUTOMATIC FROG IDENTIFICATION

Figures & data

FIGURE 1 Architecture of the intelligent frog call identifier.

FIGURE 2 A backpropagation neural network.

FIGURE 3 The web interface of the online frog recognition system.

FIGURE 4 Recognition result.

TABLE 1 Frog Species for Family Microhylidae and Phacophoridae

Family	Scientific name	Common name
Microhylidae	Microhyla heymonsi	Heymonsi narrow-mouthed toad
	Kaloula pulchra	Malaysian narrow-mouthed toad
	Microhyla butleri	Butler's rice frog
	Microhyla ornate	Ornate rice frog
	Microhyla steinegeri	Stejneger's paddy frog
Phacophoridae	Buergeria japonica	Japanese tree frog
	Buergeria robusta	Brown tree frog
	Chirixalus eiffingeri	Eiffingers tree frog
	Chirixalus idiootocus	Meintein tree frog
	Polypedates megacephalus	White-lipped tree frog
	Rhacophorus arvalis	Rhacophorus arvalis
	Rhacophorus aurantiventris	Orange-belly green tree frog
	Rhacophorus moltrechti	Moltrechtis green tree frog
	Rhacophorus prasinatus	Emerald green tree frog
	Rhacophorus taipeianus	Taipei green tree frog

TABLE 2 Number of Syllables

	Syllable
Family	Training	Testing	Total
Microhylidae	483	251	734
Phacophoridae	918	428	1346
Total	1401	679	2080

TABLE 3 Accuracy Rates ofkNN, BPNN, and NB Classifiers with Different Combination of the Input Parameters

	Classifier
	kNN		BPNN		NB
	Accuracy	Average	Accuracy	Average	Accuracy	Average
Input parameters	rate	cost	rate	cost	rate	cost
MFCCs	0.88365	1.407769	0.85272	1.365128	0.76878	1.257871
A set of seven features	0.93814	2.391338	0.88218	2.529845	0.86009	2.335718