A Two-Level Sound Classification Platform for Environmental Monitoring

Table 4

Summary of classification performance of neural network with selected features: second-level classification, two-step implementation.

	Scalar features	MFCCs	Spectrogram coefficients	Scalar features + MFCCs	Scalar features + spectrogram coefficients	MFCCs + spectrogram coefficients	Scalar features + MFCCs + spectrogram coefficients
Average correct classifications	94.33%	87.80%	80.69%	96.73%	94.27%	86.99%	94.27%
Standard deviation of correct classifications	14.98%	11.12%	27.56%	16.96%	19.64%	26.92%	19.94%
Number of features/hidden layer nodes	12/18/18	26/42/40	64/85/100	38/40/70	76/75/140	90/65/135	102/140/150