Table 3. Comparisons of effectiveness of different methods and algorithms used in pig sound and cough analysis, covering aspects, such as recognition accuracy, classification performance, and real-time detection capabilities

Detection/classification	Classification technique	Feature Extraction technique	Accuracy	Reference
Vocal classification	BP + GA	Short-time energy, Frequency centroid, Formant frequency, MFCC	93.20	[16]
	CNN-MobileNet V3	Fast Fourier transform (FFT), Log-mel spectrogram	97.52	[21]
	MnasNet	ACAM, VAD	94.72	[22]
	SVM, AdaBoost, BiLSTM	MFCC, PSD, CQT, SqueezeNet	91.41	[26]
	SE-DenseNet-121	MFCC, ΔMFCC, Δ2MFCC	93.80	[27]
	SVM	RMSE, MFCC, ZCR, Centroid, Flatness, Bandwidth, Chroma	96.45	[81]
	CNN, SVM, KNN	DNS	96.57	[49]
	TransformerCNN	MLMC	96.05	[69]

BP, back propagation; GA, genetic algorithm; MFCC, mel-frequency cepstral coefficients; CNN, convolutional neural networks; ACAM, adaptive context attachment model; VAD, voice activity detection; SVM, support vector machines; PSD, power spectral density; CQT, constant-Q transform; BiLSTM, bidirectional long short-term memory; SE, Squeeze-and-Excitation; RMSE, root-mean-square energy; ZCR, zero-crossing rates; KNN, k-nearest neighbor; DNS, dominant neighborhood structure.