Speech Synthesis Phase Correction for Natural Prosody

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing text-to-speech synthesis systems, particularly those using the TD-PSOLA model, face issues with pitch and duration modifications leading to unnatural sounding speech, phase mismatches, and metallic artifacts due to limitations in handling prosodic variations and phase information.

Innovation Solution

The method involves determining a speaker-dependent phase difference between the speech signal and its first harmonic, using this phase difference to create a window function for diphone samples, which are then concatenated to preserve natural phase information and synchronize pitch periods, thereby avoiding phase discontinuities and enhancing synthesis quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If pitch modifications are applied in TD-PSOLA model, then pitch contour is adjusted, but phase mismatches and unnatural sounding speech occur

Engineering Contradiction:

Improvepitch contour adjustmentVSAvoidphase alignment

Core Design Contradiction:

Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The patent applies preliminary action by pre-calculating and storing phase correction values in a lookup table before synthesis. The phase correction amount is determined based on the ratio of pitch periods and stored for later retrieval, avoiding real-time complex calculations and ensuring accurate phase alignment during pitch modification.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent changes the parameter of phase correction amount based on the pitch period ratio. By adjusting the phase correction parameter according to the specific pitch modification ratio being applied, the system maintains phase coherence across different pitch contours while avoiding phase mismatches.

Inventive Principle:

Principle #35Parameter changes

2Duration of action of moving object

If duration modification is applied by deleting or replicating windowed segments, then speech duration is adjusted, but quantized duration control and metallic artifacts are introduced

Engineering Contradiction:

Improvespeech durationVSAvoidduration control precision

Core Design Contradiction:

Duration of action of moving objectVSManufacturing precision

Solution Approach 1:

The patent changes the approach from quantized segment replication to continuous phase-based duration control. By using phase correction amounts derived from pitch period ratios, the system achieves fine-grained duration control without the one-pitch-period quantization limitation, eliminating metallic artifacts while maintaining natural speech duration.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If pitch-synchronous overlap-add is used, then pitch period modification is achieved, but compensation for duration modification is required

Engineering Contradiction:

Improvepitch period modificationVSAvoidcompensation processing

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges pitch period modification and duration modification into a single unified process. By calculating phase correction amounts based on the ratio of pitch periods, the system simultaneously achieves both pitch and duration adjustments without requiring separate compensation steps, thereby reducing overall system complexity.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS7822599B2Method for synthesizing speech

Publication Date: 2010.10.26 HUAWEI TECH CO LTD

AI summary

The present invention relates to a method for analyzing speech, the method comprising the steps of: a) inputting a speech signal, b) obtaining the first harmonic of the speech signal, c) determining the phase-difference Df between the speech signal and the first harmonic.