To enhance the communication between human and robots at home in the future, speech synthesis interfaces are indispensable that can generate expressive speech. In addition, synthesizing celebrity voice is commercially...To enhance the communication between human and robots at home in the future, speech synthesis interfaces are indispensable that can generate expressive speech. In addition, synthesizing celebrity voice is commercially important. For these issues, this paper proposes techniques for synthesizing natural-sounding speech that has a rich prosodic personality using a limited amount of data in a text-to-speech (TTS) system. As a target speaker, we chose a well-known prime minister of Japan, Shinzo Abe, who has a good prosodic personality in his speeches. To synthesize natural-sounding and prosodically rich speech, accurate phrasing, robust duration prediction, and rich intonation modeling are important. For these purpose, we propose pause position prediction based on conditional random fields (CRFs), phone-duration prediction using random forests, and mora-based emphasis context labeling. We examine the effectiveness of the above techniques through objective and subjective evaluations.展开更多
A new approach for reducing error of the volume penalization method is proposed.The mask function is modified by shifting the interface between solid and fluid by√νηtoward the fluid region,whereνandηare the visco...A new approach for reducing error of the volume penalization method is proposed.The mask function is modified by shifting the interface between solid and fluid by√νηtoward the fluid region,whereνandηare the viscosity and the permeability,respectively.The shift length√νηis derived from the analytical solution of the one-dimensional diffusion equation with a penalization term.The effect of the error reduction is verified numerically for the one-dimensional diffusion equation,Burgers’equation,and the two-dimensional Navier-Stokes equations.The results show that the numerical error is reduced except in the vicinity of the interface showing overall second-order accuracy,while it converges to a non-zero constant value as the number of grid points increases for the original mask function.However,the new approach is effective when the grid resolution is sufficiently high so that the boundary layer,whose width is proportional to√νη,is resolved.Hence,the approach should be used when an appropriate combination ofνandηis chosen with a given numerical grid.展开更多
文摘To enhance the communication between human and robots at home in the future, speech synthesis interfaces are indispensable that can generate expressive speech. In addition, synthesizing celebrity voice is commercially important. For these issues, this paper proposes techniques for synthesizing natural-sounding speech that has a rich prosodic personality using a limited amount of data in a text-to-speech (TTS) system. As a target speaker, we chose a well-known prime minister of Japan, Shinzo Abe, who has a good prosodic personality in his speeches. To synthesize natural-sounding and prosodically rich speech, accurate phrasing, robust duration prediction, and rich intonation modeling are important. For these purpose, we propose pause position prediction based on conditional random fields (CRFs), phone-duration prediction using random forests, and mora-based emphasis context labeling. We examine the effectiveness of the above techniques through objective and subjective evaluations.
文摘A new approach for reducing error of the volume penalization method is proposed.The mask function is modified by shifting the interface between solid and fluid by√νηtoward the fluid region,whereνandηare the viscosity and the permeability,respectively.The shift length√νηis derived from the analytical solution of the one-dimensional diffusion equation with a penalization term.The effect of the error reduction is verified numerically for the one-dimensional diffusion equation,Burgers’equation,and the two-dimensional Navier-Stokes equations.The results show that the numerical error is reduced except in the vicinity of the interface showing overall second-order accuracy,while it converges to a non-zero constant value as the number of grid points increases for the original mask function.However,the new approach is effective when the grid resolution is sufficiently high so that the boundary layer,whose width is proportional to√νη,is resolved.Hence,the approach should be used when an appropriate combination ofνandηis chosen with a given numerical grid.
