Gait profiling for Identikit ®

Facial composite produced by FACES 3.0 software, similar to the Identi-kit software. Photo:

Facial composite produced by FACES 3.0 software, similar to the Identi-kit software. Photo:

Identikit ® is well known as a system for developing facial composite sketches of criminal suspects based on the verbal reports of witnesses. The system has been remarkably successful, but it is nonetheless limited in the kinds of suspect characteristics that it utilizes. In particular, all of the characteristics that are captured by an Identikit ® compostite are static. For example, we might learn something about the shape of a suspect’s lips from a compostite, but we would know nothing about how the those lips move when the suspect speaks.

In the past couple of decades, vision scientists have made considerable advances in understanding how the visual system processes motion. Of particular interest is the fact that the visual systems of both humans and other species (e.g., cats) are capable of special processing of motion information that derives from a biological (usually conspecific) sources (see, for example, Troje, N. F., 2002, Decomposing biological motion: A framework for analysis and synthesis of human gait patterns. Journal of Vision, Vol. 2, No. 5, 371-387).

Using the models of human gait developed by scientists like Troje, one can imagine a system for profiling the appearance of a suspect’s walk. This might be done by having a witness view and pairs of models showing two different gaits with the object of having the witness choose which of the two is most like that of the suspect. Alternatively, a process of adjustment, in which each of the gait-relevant variables was altered, could be used. Other psychophysical methods would lead to related solutions.

The output from a witness’s profile of a the gait of a suspect might be used to help in the apprehension of the suspect. For example, even if a suspect were unable to articulate beliefs about the suspect’s age or sex,information about those two variables might be derivable from a model of the suspect’s gait which the witness had helped to build. Naturally, if such a system were successfully developed, there would be no reason to limit consideration to gait. Arm movements, gestures, and other biological motion could be included.

Contributors: Mark R. Diamond