 As part of our effort to develop authoring tools for designers of Equator-like experiences we applied the concept of learning-by-example. To demonstrate this technique we used two dodecahedra with a different animal or number on each face. Inside each dodecahedron we placed a Smart-IT sensor module hosting a 3-axis accelerometer (2x2-axis accelerometers) with wireless RF communications. When static, the accelerometers measure the force of the gravitational field acting on each of the 3 axes. This information can be used to determine the orientation of the dodecahedron, and therefore which side is on the top.  However, it is our aim that the authoring tools should be useable by non-technical designers. Therefore, rather than having to calculate the geometry directly, learn-by-example allows the designer to simply “show” the system a set of examples for identification. In this particular case, the dodecahedron is rotated such that each side is facing up and then assigned a label, such as “cat”, “dog”, “3”, or “Three”. Behind the scenes, the raw accelerometer data is collected and used to train a neural-network classifier to identify each of the labels. The designer can then use these labels in the authoring environment, rather than having to handle the raw accelerometer values directly. |