*2.3. Fusion Module*

The feature representations obtained from both fingerprint and ECG branches are further processed using a fusion module. This fusion module is composed of a sequence of layers, as shown in Figure 7. First, the feature vector of an ECG is concatenated into a fingerprint feature vector to produce a single global feature vector. The concatenated feature is fed to an additional fully connected layer followed by a BN, Swish activation function, and dropout (0.5) regularization. Finally, the output of this module is fed to a binary classification layer using a sigmoid activation function to determine the final fingerprint class, i.e., an artefact or bona fide.

**Figure 7.** Structure of the fusion module.
