Abstract
We address the problem of retrieving an object description, in the form of a silhouette, from a database of shapes, in a way that is robust with respect to translation, rotation, scale change, and distortion. Algorithms to solve this problem can be a key step in passive assistant software that receives information from available sensors and databases, and provides useful information about the local environment to a visually disabled user (where the assistant input may be in the form of audio). Experiments show significant gains in retrieval accuracy over previous work.