Investigation of Efficiency and Observation of Voice Quality Parameters in Ideal User’s Voice Creation using UTAU and Interactive Differential Evolution

Abstract

Many papers and books on improving voice quality exist around the world. This suggests that some people want to change their voice. However, there is a difficult image on improving voice quality. The reason is that objectively evaluating your own voice training is difficult. In this study, to improve the voice quality based on the user's voice, we propose a method to create ideal voice heard from other's using software UTAU and Interactive Differential Evolution. As result of the search and evaluation experiments, the evaluation values significantly increased from the initial to the final generations. These results showed that the proposed method has the possibility of making the ideal voice of the user. Furthermore, overall convergence was observed when in the transition of individual parameters. Especially, in the transition of the parameter related to voice formant, obvious convergence was confirmed in 13 out of all 14 subjects.

Making an English Speech Similar to the User’s Voice using UTAU and Interactive Differential Evolution

“…Paired comparison-based IDE, which is known as an efficient and faster way to evaluate, is used [16]. The paired comparison-based IDE was used in the previous studies to create content related to a sense of hearing [12,13]. Furthermore, the paired comparisonbased IDE was also used in the creation of content related to other senses [19,20].…”

“…Originally, UTAU is used for creating singing voices as the name says (UTAU comes from a Japanese verb "歌う" that means "sing"). Some studies used UTAU [11,12]; however, this area is undeveloped. Figure 2 is a screenshot of UTAU.…”

“…Interactive Evolutionary Computation (IEC) was known as a method that obtains the user's media content based on a stochastic search [6,7], and previous studies created sound content using IECs [8][9][10][11][12][13][14]. The purpose of this study is to propose a method to produce the voice data of the model pronunciation similar to the user's voice by using UTAU [15] and interactive differential evolution (IDE) [16], one of the IECs.…”