Adaptive Interpolation Factor Selection for Speech Synthesis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing electronic devices face challenges in accurately decoding audio signals due to losses during transmission, leading to suboptimal quality of reconstructed speech, particularly when frame erasures occur, as they struggle to adaptively handle changes in speech characteristics.

Innovation Solution

A method for determining an interpolation factor set based on the energy ratio between current and previous frames, which involves interpolating subframe line spectral frequency (LSF) vectors to synthesize speech signals, using a prediction mode indicator to decide between predictive and non-predictive quantization, and employing adaptive interpolation factors to mitigate artifacts from frame erasures.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If audio signals are encoded to reduce bandwidth, then transmission efficiency is improved, but decoding accuracy deteriorates when frame erasures occur

Engineering Contradiction:
Improvebandwidth consumptionVSAvoiddecoding accuracy
Core Design Contradiction:
Loss of energyVSMeasurement precision

Solution Approach 1:

The patent implements dynamic interpolation factor selection that adapts to changing speech conditions. The decoder calculates energy ratios between consecutive frames and selects interpolation factors from different sets (first set for normal conditions, second set for frame erasures) based on whether the energy ratio falls within a predetermined range. This dynamic adaptation allows the system to maintain high decoding accuracy under varying transmission conditions while keeping bandwidth consumption low.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the interpolation parameter based on speech characteristics. By calculating the energy ratio of consecutive frames and comparing it to a threshold, the system selects different interpolation factor sets. When the energy ratio indicates normal speech conditions, one set of interpolation factors is used; when it indicates frame erasures, a different set is applied. This parameter change strategy resolves the contradiction by optimizing decoding accuracy for specific conditions without increasing overall bandwidth consumption.

Inventive Principle:
Principle #35Parameter changes

2Device complexity

If fixed interpolation factors are used for LSF vector interpolation, then device complexity is reduced, but speech quality deteriorates under varying speech conditions

Engineering Contradiction:
Improveinterpolation control complexityVSAvoidspeech synthesis quality
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent transitions from fixed to dynamic interpolation factor selection. The system calculates energy ratios between consecutive speech frames and dynamically selects interpolation factors from appropriate sets based on whether the ratio indicates normal conditions or frame erasures. This dynamic approach maintains speech synthesis quality under varying conditions while keeping the control mechanism relatively simple through predetermined factor sets and threshold-based selection.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The decoder performs self-adaptation by automatically detecting frame erasures through energy ratio calculation and selecting appropriate interpolation factors without external control. The system uses its own decoded speech energy measurements to determine when to switch between different interpolation factor sets, enabling quality maintenance under varying conditions without adding complex external control mechanisms.

Inventive Principle:
Principle #25Self-service

3Manufacturing precision

If adaptive interpolation factors are selected based on energy ratio, then speech synthesis quality is improved, but computational complexity increases

Engineering Contradiction:
Improvespeech synthesis qualityVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent implements a dynamic but computationally efficient adaptation mechanism. The system calculates simple energy ratios between consecutive frames and uses threshold comparison to select from predetermined interpolation factor sets. This approach provides adaptive speech synthesis quality improvement while maintaining low computational complexity through the use of pre-calculated factor sets and simple threshold-based decision logic.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent performs preliminary preparation by establishing predetermined interpolation factor sets and energy ratio thresholds before actual speech decoding. This allows the runtime system to simply compare measured energy ratios against pre-determined thresholds and select from pre-prepared factor sets, significantly reducing computational complexity while maintaining adaptive speech synthesis quality.

Inventive Principle:
Principle #10Preliminary action

4Device complexity

If simple energy ratio comparison is used for mode selection, then device complexity is reduced, but adaptability to speech changes deteriorates

Engineering Contradiction:
Improvecontrol mechanism complexityVSAvoidadaptability to speech characteristics
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent achieves dynamic adaptation to speech characteristics through energy ratio comparison. By calculating the ratio of energies between consecutive frames and comparing it to a predetermined threshold, the system adapts its interpolation behavior to match current speech conditions. This simple yet effective mechanism provides good adaptability while maintaining low device complexity.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system uses its own decoded speech energy measurements to automatically detect frame erasures and select appropriate interpolation factors. This self-service approach provides adaptability to speech changes without requiring external control or complex analysis, achieving a balance between simplicity and responsiveness to speech characteristics.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP2959483B1Systems and methods for determining an interpolation factor set
Publication Date: 2017.12.27 QUALCOMM INC
  • EP2959483B1 patent drawingFigure 1
  • EP2959483B1 patent drawingFigure 2
  • EP2959483B1 patent drawingFigure 3

AI summary

A method for determining an interpolation factor set by an electronic device is described. The method includes determining a value based on a current frame property and a previous frame property. The method also includes determining whether the value is outside of a range. The method further includes determining an interpolation factor set based on the value and a prediction mode indicator if the value is outside of the range. The method additionally includes synthesizing a speech signal.