Interactive Evolutionary Computation Improving Voice Impressions with Keeping Speaker Personality for Real-Time Speech

Recently, we generally have meetings via the Internet. In this situation, we use background display to improve our impression of other members of the meetings. To improve the users’ voice via the Internet, this study proposes an Interactive Evolutionary Computation (IEC) that adjusts the voice filter based on real-time pronunciations while keeping user’s personality. The concrete system was constructed by employing a Genetic Algorithm and Koigoe, a software voice filter. The listening experiments were conducted to investigate the efficiencies of the proposed IEC from perspectives of increasing the fitness values and keeping the speaker’s personality. The results showed that the proposed IEC has enough possibility to find a good parameter set of the voice filter; however, we need to improve its performance because the obtained best filter did not overcome the impression of the original voice without any filter. Furthermore, the proposed IEC could be considered to keep the user’s personality based on the result of the evaluation experiment.