Voice Synthesis Apparatus Pitch Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing voice synthesis technologies face challenges in creating natural-sounding voices when interpolating phoneme piece data across different pitches, leading to unnatural tone transitions and acoustic characteristics.
Innovation Solution
A voice synthesis apparatus that interpolates between phoneme piece data of different sound characteristics, such as pitch, using a phoneme piece interpolation part to create target phoneme piece data by adjusting spectra and sound volumes, while considering the distinction between voiced and unvoiced sounds, and employing a continuant sound interpolation part for fluctuation components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If phoneme piece data of representative pitches are adjusted to create new phoneme piece data of target pitch, then phoneme piece data for all pitch levels can be generated from limited original data, but the tones of synthesized sounds with adjacent pitches become dissimilar and unnatural
Solution Approach 1:
The patent introduces an intermediary phoneme piece data structure that serves as a bridge between original phoneme piece data of different pitches. By creating intermediate phoneme piece data through spectral interpolation between original data of different pitches (e.g., between E3 and G3 to create F3 and F#3), the system ensures smooth tone transitions while maintaining natural acoustic characteristics. The intermediary data contains spectrums that are mathematically interpolated between the original spectrums, ensuring continuity and naturalness.
2Ease of manufacture
If phoneme piece data are prepared with respect to several representative pitches, then data preparation workload is reduced, but the quality of synthesized sounds at non-representative pitches deteriorates due to tone discontinuity
Solution Approach 1:
The patent performs preliminary interpolation to create intermediate phoneme piece data before actual voice synthesis. By pre-calculating and storing phoneme piece data for intermediate pitches (F3, F#3) through spectral interpolation of original data (E3, G3), the system ensures that when synthesis is needed, natural-sounding data is already prepared. This preliminary action eliminates the need for real-time interpolation during synthesis, ensuring both quality and efficiency.
3Adaptability or versatility
If pitch adjustment is applied to create synthesized sounds at different pitches, then voice synthesis can be performed with limited original data, but abrupt tone changes occur at transition points between adjacent pitches
Solution Approach 1:
The patent applies parameter changes to the spectral parameters of phoneme piece data through mathematical interpolation. By systematically varying spectral parameters (amplitude, frequency distribution) between original phoneme piece data of different pitches, the system creates a continuous parameter space where intermediate pitches have intermediate spectral characteristics. This ensures that transition points between adjacent pitches (e.g., F3 to F#3) show gradual parameter changes rather than abrupt jumps, maintaining tone stability.
Data Source
AI summary
In a voice synthesis apparatus, a phoneme piece interpolator acquires first phoneme piece data corresponding to a first value of sound characteristic, and second phoneme piece data corresponding to a second value of the sound characteristic. The first and second phoneme piece data indicate a spectrum of each frame of a phoneme piece. The phoneme piece interpolator interpolates between each frame of the first phoneme piece data and each frame of the second phoneme piece data so as to create phoneme piece data of the phoneme piece corresponding to a target value of the sound characteristic which is different from either of the first and second values of the sound characteristic. A voice synthesizer generates a voice signal having the target value of the sound characteristic based on the created phoneme piece data.