Customising a qualitative colour description for adaptability and usability

A method for adapting a colour naming system to specific groups of people is presented.The starting point is a reference parametrisation of a qualitative colour description (QCD) model.The method applies a classifier over data gathered from the target group.The adapted model significantly improves the suitability of the colour naming system for the target group.The elementary colours in the refined QCD wheel are compared to other colour wheels in the literature. Colour naming consists of successfully finding the correspondence between colours as named by humans and the colour coordinates used by machine displays. Its successful implementation is crucial for human-machine interaction tasks, e.g. for the communication between service robots and humans. However, significant variability among human groups makes a general solution to this problem notoriously difficult. Qualitative models are appropriate in this context because they are robust in the presence of such variability. This paper contributes an approach to adapt the qualitative colour description (QCD) model to specific groups of users by gathering experimental data from them using a simple web interface. A classifier based on the geometric mean statistic is applied to adjust the correspondence between colour names and HSL coordinates in order to further approximate the QCD model to the common human colour understanding within the group. The usability of the resulting adapted model is evaluated by naming the predominant colour in a set of images extracted from Google searches. Results show that the refined QCD model can be used successfully to provide reference and grounding in human-machine communication. Finally, the elementary colours in the refined QCD wheel are also compared to other colour wheels in the literature.