PSI - Issue 64

Sasan Farhadi et al. / Procedia Structural Integrity 64 (2024) 549–556

552

4

S. Farhadi et al. / Structural Integrity Procedia 00 (2024) 000–000

Waveform sampling rate = 100 kHz time duration = 0.0102 s

Framing no. frames = 384 window function: Hamming

Discrete Fourier Transform (DFT) no. FTT = window length

Warping the Freq. to Mel-scale transformation method: Mel

no. mel bands = 32 filter type: triangular

Discrete Cosine Transform (DCT) DCT type: type-III

MFCCs no. bands = 32 MFCC shape = (32, 128)

Fig. 1: Process to extract MFCC features

Fig. 2: Activations functions and their derivatives

Moreover, it is worth noting that activation functions play a crucial role in neural networks by determining the output of neural processing units 2. These functions and their derivatives are crucial for the backpropagation process, allowing the model to adjust its parameters correctly during training.