Neural Vocoder Waveform Reconstruction for Higher-Quality Audio

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing artificial intelligence models struggle to produce high-quality waveform data due to limitations in model complexity, leading to suboptimal performance in tasks requiring complex processing.

Innovation Solution

An electronic device employs a neural vocoder comprising an encoder and decoder model to improve waveform data quality by training the encoder to output feature data based on input acoustic data, and the decoder to generate high-quality waveform data by minimizing errors between estimated and actual waveform data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a touch-sensitive display is used to enable users to view and control media playback, then ease of operation is improved, but the display may be damaged by pressure applied during touch input

Engineering Contradiction:
Improveease of operationVSAvoiddisplay durability
Core Design Contradiction:
Ease of operationVSStrength

Solution Approach 1:

A flexible membrane with piezoelectric elements is introduced as an intermediary between the touch-sensitive display and the user's finger. This membrane allows touch input to be detected while protecting the display from direct pressure damage. The piezoelectric elements convert mechanical pressure into electrical signals, enabling touch detection without requiring the display itself to withstand the mechanical stress.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If multiple sensors are used to detect finger position and pressure, then measurement precision is improved, but device complexity increases

Engineering Contradiction:
Improvemeasurement precisionVSAvoiddevice complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

Multiple functions are merged into a single integrated membrane structure. The flexible membrane incorporates both position detection capabilities through its flexibility and pressure detection through embedded piezoelectric elements. This unified approach allows simultaneous detection of finger position and applied pressure while avoiding the complexity of separate sensor systems.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The flexible membrane serves multiple functions simultaneously: it acts as a protective layer for the display, a position sensor through its deformation characteristics, and a pressure sensor through the piezoelectric elements. This multi-functional design reduces the need for separate components and simplifies the overall device architecture.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

The approach enables the production of waveform data with enhanced sound quality by refining the decoder model's performance through training with error-based adjustments, resulting in improved output quality.

Implementation Method 1

the membrane may include a piezoelectric film that generates an electrical signal in response to pressure applied to the membrane

Methodology Applied
Scientific EffectPiezoelectric effect: Piezoelectric Effect

Data Source

PatentEP4207192B1Electronic device and method for controlling same
Publication Date: 2026.05.06 SAMSUNG ELECTRONICS CO LTD
  • EP4207192B1 patent drawingFigure 1
  • EP4207192B1 patent drawingFigure 2
  • EP4207192B1 patent drawingFigure 3

AI summary

The disclosure relates to an electronic device and a control method thereof. The electronic device includes a memory, and a processor configured to: obtain first feature data for estimating a waveform by inputting acoustic data of a first quality to a first encoder model; and obtain waveform data of a second quality that is a higher quality than the first quality by inputting the first feature data to a decoder model to.