The modal properties of the shear frame were computed using analytical approach for a comparison with the experimental modal frequencies.
Then, the values of the motion parameters (rotations and translations) between {y k+1} k=1,2,…,M and y 1 (reference frame) are computed using the Keren et al. algorithm which is chosen for its reliability, robustness, accuracy, and computational efficiency.
Thus, the mask at frame n ∘ + 1 is computed using the predicted parameters from the corrected partials at frame n ∘.
At the mth frame, the LSFM feature L x (m) is computed using the previous R frames.
The features in this case were computed using 25 ms frames with 10 ms of frame shifts.
The results are computed using 8 frames and 5 wavelet decomposition levels.
The estimated position of the abnormality in the current frame, ( μ x, μ y ), is computed using the affine transformation of the previous abnormality's position (t x,t y ), i.e., left[begin{array}{c} mu_{x} mu_{y}1 end{array} right] = mathbf{A_{text{textit{s,t,u}}}} left[ begin{array}{c} t_{x} t_{y} 1end{array} right].
Then, at each frame t, a new smoothed score function S ̃ ( C i ′ ( t ) ) for a combination C i ′ ( t ) is computed using the neighbor frames: S ̃ ( C i ′ ( t ) ) = ∑ j = t - K t + K S ( C ′ i ( j ) ) (11).
For CNNs, the log power spectrogram was computed using 10 ms frames with a 10 ms shift, which performed the best.
The STFT was computed using 256 samples long frames, having an overlap of 128 samples between successive frames (50%% overlap).
