This web-demo illustrates an approach to dimensional affect recognition. Targets (arousal and valence) and features from heart rate variability are provided for training and testing. Click on “Submit” button (with default configuration and pre-loaded data) to reproduce the paper figures.
Two methods are compared. The first one is a Supervised Self-Organization Map (sSOM), that can be used for recognition as well as for analyzing patterns in data. Some sSOM parameters are:
- Map size: number of neural units [height x width].
- Lambda (λ): Weight factor applied to the targets. It change topological ordering.
- Features to show: Indicates the features to be shown in the map (default: “1 7 21 25 2”)
The second method is the kernel version of Extreme Learning Machines (kELM). kELM parameters are:
- C: Regulation coefficient.
- Gamma (γ): Kernel’s exponential coefficient.
Contact: Leandro Bugnon