Human-centered interfaces (2)
Data-level fusion implies the concatenation of data possessing high level of synchronization before feature extraction.
Feature fusion implies the concatenation of features provided by the different modalities.
- When feature fusion aims at providing a combined fusion
output, Kalman filters are usually exploited.
- When feature fusion is integrated with the decision maker,
HMM’s or multilayer perceptrons are employed.
Most commonly used type of fusion is decision-level fusion which is more robust to individual modality failure.