Abstract: The way music is interpreted by the human brain is a very interesting topic, but also an intricate one. Although this domain has been studied for over a century, many gray areas remain in the understanding of music. Recent advances have enabled us to perform accurate measurements of the time taken by the human brain to interpret and assimilate a sound. Cognitive computing provides tools and development environments that facilitate human cognition simulation. ACT-R is a cognitive architecture which offers an environment for implementing human cognitive tasks. This project combines our understanding of the music interpretation by a human listener and the ACT-R cognitive architecture to build SINGER, a computerized simulation for listening and recalling songs. The results are similar to human experimental data. Simulation results also show how it is easier to remember short melodies than long melodies which require more trials to be recalled correctly.
Abstract: Cognitive models allow predicting some aspects of utility
and usability of human machine interfaces (HMI), and simulating
the interaction with these interfaces. The action of predicting is based
on a task analysis, which investigates what a user is required to do
in terms of actions and cognitive processes to achieve a task. Task
analysis facilitates the understanding of the system-s functionalities.
Cognitive models are part of the analytical approaches, that do not
associate the users during the development process of the interface.
This article presents a study about the evaluation of a human
machine interaction with a contextual assistant-s interface using ACTR
and GOMS cognitive models. The present work shows how these
techniques may be applied in the evaluation of HMI, design and
research by emphasizing firstly the task analysis and secondly the
time execution of the task. In order to validate and support our
results, an experimental study of user performance is conducted at
the DOMUS laboratory, during the interaction with the contextual
assistant-s interface. The results of our models show that the GOMS
and ACT-R models give good and excellent predictions respectively
of users performance at the task level, as well as the object level.
Therefore, the simulated results are very close to the results obtained
in the experimental study.