Conference Paper

A proposal for optimization of sign sound using interactive differential evolution.

DOI: 10.1109/ICSMC.2011.6083849 Conference: Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, Anchorage, Alaska, USA, October 9-12, 2011
Source: DBLP

ABSTRACT Interactive Evolutionary Computation (IEC) is known as an effective method to create media contents suited to user's preference and objectives to use. The present study proposed an IEC method that creates sign sounds with interactive differential evolution. In the user's evaluation, paired comparison was employed: the user selects better one from two presented media contents instead of scoring many solution candidates in a same time. To fundamentally investigate the efficacy of the proposed method, two listening experiments were performed; experiment 1 as comparing experiment and experiment 2 as re-evaluating experiment. Target of the creation with the proposed method was warning sign sounds. Eight males participated as subjects in both experiment. In the result of the experiment 1, gradual decrease of total Euclidean distance between eight DE's vectors in each generation was observed. In the result of the experiment 2, gradual increase of the subjective fitness value was observed. These results were not significant, however, shrink of searching space and increase of fitness value suggested a possibility of the proposed method to create sign sounds suited to the user.

0 Followers
 · 
56 Views
  • [Show abstract] [Hide abstract]
    ABSTRACT: Interactive Evolutionary Computation (IEC) is well known as an effective method to create media contents suited to user's preference and objectives to use. As one of the methods, we have applied Differential Evolution, which is recent evolutionary algorithm to IEC. Concretely, we have already presented a method that creates sign sounds with Interactive Differential Evolution (IDE). This study aims to investigate fundamentally the efficacy of the IDE method in comparison with Interactive Genetic Algorithm (IGA). Two listening experiments were conducted: experiment 1 as a creation experiment with IDE and IGA, experiment 2 as a re-evaluation experiment. Target of the creation was warning sign sounds. In the experiment 2, representative five sign sounds created in both of IDE and IGA were evaluated. Sixteen males participated as subjects in the experiments. In the result of the experiment 1, IDE overcame IGA in subjective fitness value. Drastic shrink of searching space was observed in IGA, and larger time cost was observed in IDE. In the result of the experiment 2, higher fitness value in average was observed in IDE, however, the difference was not significant.
    Software Engineering, Artificial Intelligence, Networking and Parallel & Distributed Computing (SNPD), 2012 13th ACIS International Conference on; 01/2012