InfoMus Lab
The EyesWeb Space Analysis Library
The EyesWeb Space Analysis Library is based on a model considering a collection of discrete potential functions defined on a 2D space. The space is divided into active cells forming a grid. A point moving in the space is considered and tracked. Three main kinds of potential functions are considered: (i) potential functions not depending on the current position of the tracked point, (ii) potential functions depending on the current position of the tracked point, (iii) potential functions depending on the definition of regions inside the space.
Objects and subjects in the space can be modeled by time-varying potentials. Regions in the space can also be defined. A certain number of "meaningful" regions (i.e., regions on which a particular focus is placed) can be defined and cues can be measured on them (e.g., how much time a tracked subject occupied a given region).
The metaphor can be applied both to real spaces (e.g., scenery and actors on a stage, the dancer's General Space as described by Rudolf Laban) and to virtual, semantic, expressive spaces (e.g., a space of parameters where gestures are represented as trajectories): for example, if, from the one hand, the tracked point is a dancer on a stage, a measure of the time duration along which the dancer was in the scope of a given light can be obtained; on the other hand, if the tracked point represents a position in a semantic, expressive space where regions corresponds to basic emotions, the time duration along which a given emotion has been recognized can also be obtained.
The EyesWeb Space Analysis Library implements the model and includes blocks allowing the definition of interacting discrete potentials on 2D spaces, the definition of regions, and the extraction of cues (such as, for example, the occupation rates of regions in the space). For example, the figure below shows the occupation rates calculated on a rectangular space divided into 25 cells. The intensity (saturation) of the color for each cell is directly proportional to the occupation rate of the cell.
 |
| Occupation rates calculated with the EyesWeb Space Analysis Library for a trajectory in a 2D space divided into 25 cells. The displayed trajectory refers to the last 25 frames (i.e., 1 s), but the occupations rates are calculated for the whole trajectory since the start of the gesture. |