Speech Quality Assessment via Perceptual Dimension Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current speech quality assessment models for communication systems, particularly for wideband and mixed-band speech transmission, lack accuracy in estimating quality scores and fail to identify the source of quality loss in transmission paths.

Innovation Solution

A full-reference signal-based model that estimates speech quality by determining perceptual dimensions such as continuity, noisiness, and loudness through pre-processing and analyzing the energy gradient, musical tones, and frequency spectrum of speech signals, integrating these dimensions to provide a comprehensive quality measure.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If full-reference signal-based models are used to estimate speech quality, then measurement precision is improved, but the ability to identify the source of quality loss deteriorates

Engineering Contradiction:
Improvequality score estimation accuracyVSAvoidsource identification capability
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent segments the speech quality assessment into multiple independent perceptual dimensions (continuity, noisiness, loudness, frequency content) rather than providing a single aggregate score. Each dimension is calculated separately through dedicated analysis modules, allowing both precise measurement of overall quality and identification of specific degradation sources through dimension profiling.

Inventive Principle:
Principle #1Segmentation

2Reliability

If subjective tests with human subjects are conducted, then reliability of quality assessment is improved, but productivity deteriorates

Engineering Contradiction:
Improvequality assessment reliabilityVSAvoidassessment speed
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent replaces the mechanical system of human subject testing with an automated signal processing system that analyzes speech signals through perceptual models. The system uses algorithms to compute perceptual dimensions (continuity, noisiness, loudness) directly from signal characteristics, eliminating the need for human listeners while maintaining reliability through psychoacoustically validated measurement methods.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Loss of information

If detailed perceptual dimension analysis is performed, then loss of information is reduced, but device complexity increases

Engineering Contradiction:
Improvequality detail informationVSAvoidsystem complexity
Core Design Contradiction:
Loss of informationVSDevice complexity

Solution Approach 1:

The patent divides the complex task of speech quality assessment into separate functional modules, each responsible for calculating a specific perceptual dimension (continuity, noisiness, loudness, frequency content). This modular segmentation reduces device complexity by allowing independent implementation and optimization of each dimension's analysis algorithm while preserving comprehensive quality information.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP2037449B1Method and system for the integral and diagnostic assessment of listening speech quality
Publication Date: 2017.11.01 ORANGE SA
  • EP2037449B1 patent drawingFigure 1
  • EP2037449B1 patent drawingFigure 2
  • EP2037449B1 patent drawing

AI summary

In order to determine a speech quality measure related to a signal path of a data transmission system utilized for speech transmission the invention proposes methods for determining a speech quality measure of an output speech signal (y) with respect to an input speech signal (x), wherein said input signal (x) passes through a signal path (100) of a data transmission system resulting in said output signal (y). The invention further proposes respective devices and a system adapted to perform the respective methods. The characteristics of the inventive approach comprise an estimation of individual perceptually-motivated dimension scores with the help of dedicated estimators, integration of a basic listening quality score obtained with the help of a full-reference model and the dimension scores into an overall quality estimation, and separate output of the overall quality score and the dimension scores for the purpose of planning, designing, optimizing, implementing, analyzing and monitoring speech quality.