Lightweight feature extraction method for efficient acoustic-based animal recognition in wireless acoustic sensor networks

EURASIP Journal on Wireless Communications and Networking

Table 4 WASN-based animal recognition systems

	Approach	Features	Number of features	Classifier	Dataset (no. of classes) = no of files	Recognition accuracy (Recall)	Sampling frequency
1	Croker et al. [11]	Frequency and time	5	ED	Frog (5) = 100	Recall = 85% Accuracy = 89%	16 kHz
2	Dang et al. [12]	Envelope Extraction	Not specified	Matched filtering	Frog (3) = not specified	Accuracy = 90%	< 10 kHz
3	Wei et al. [13]	From Gradient Projection for Sparse Reconstruction	featureless using a sparse representation	Their own \({\varvec{\iota}}\)₁-minimization Sparse Approximation-based classifier	Frog (14) = 228	Recall \(\approx\) 98%	24 kHz
3	Wei et al. [13]	From Gradient Projection for Sparse Reconstruction	featureless using a sparse representation		crickets (20) = 663	Recall \(\approx\) 50%	24 kHz
4	Colonna et al. [10]	Wavelet	4	k-NN	Anurans(9) = 49 syllables	96.25% 94.16 86.96%	44.1 kHz 11 kHz 5.5 kHz
5	Algobail et al. [19]	Time	2	ED	Animals (7) = 114	81.34%	44.1 kHz
6	Our scheme	Wavelet	2	MD ED	Animals (12) = 587	85.59% 86.06%	8 kHz