Speech Correlation Structure for Neurological Assessment

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for evaluating Major Depressive Disorder (MDD) rely on questionnaire-based assessments like the Hamilton Depression Rating Scale and Beck Depression Inventory, which have validity and reliability concerns, necessitating a more objective and reliable evaluation method.

Innovation Solution

A method and system that process speech-related variables to extract channel-delay correlation structures, allowing for the generation of assessments independent of language and noise, using features like formant frequencies, Mel Frequency Cepstral Coefficients, and Delta Mel Frequency Cepstral Coefficients to assess conditions such as MDD.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If questionnaire-based assessment tools (HAMD, BDI) are used to evaluate MDD, then clinical assessment can be obtained, but validity and reliability concerns arise

Engineering Contradiction:
Improveassessment validityVSAvoidassessment reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent replaces the mechanical/questionnaire-based assessment system with an acoustic/speech analysis system. Instead of using self-report questionnaires (HAMD, BDI), the invention extracts acoustic features from speech signals and analyzes their correlation structures to assess MDD, thereby substituting subjective questionnaire methods with objective acoustic measurement methods.

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

Solution Approach 2:

The patent changes the measurement parameters from questionnaire scores to acoustic feature correlations. By analyzing the correlation structure of acoustic features (such as formant frequencies, spectral characteristics) in speech signals, the system transforms the assessment parameters from subjective self-report scores to objective acoustic measurement parameters, improving both validity and reliability.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If speech processing is used to extract vocal tract representations, then language-independent assessment is achieved, but processing complexity increases

Engineering Contradiction:
Improvelanguage independenceVSAvoidprocessing complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent extracts specific acoustic features from speech signals that are independent of linguistic content. By focusing on extractions of vocal tract characteristics (formant frequencies, spectral correlations) rather than linguistic meaning, the system achieves language independence while managing processing complexity through selective feature extraction.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the speech signal into distinct acoustic features (formant frequencies, spectral characteristics, correlation coefficients). By dividing the complex speech processing task into separate feature extraction and correlation analysis stages, the system achieves language independence through modular processing, reducing overall complexity.

Inventive Principle:
Principle #1Segmentation

3Reliability

If channel-delay correlation structure is extracted from vocal tract representations, then noise-resistant assessment is achieved, but computational requirements increase

Engineering Contradiction:
Improvenoise resistanceVSAvoidcomputational energy
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent changes the analysis parameters from raw acoustic signals to correlation structures. By computing channel-delay correlation coefficients that capture the temporal and spectral relationships in speech, the system transforms noisy raw signals into robust correlation parameters that are resistant to noise while requiring computational energy only for the correlation calculation.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent performs preliminary extraction of vocal tract representations before conducting correlation analysis. By pre-processing the speech signal to extract relevant acoustic features and their correlations in advance, the system prepares noise-resistant parameters that can be used for assessment without requiring intensive real-time computational energy during the actual evaluation.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3057493B1Using correlation structure of speech dynamics to detect neurological changes
Publication Date: 2020.06.24 MASSACHUSETTS INST OF TECH
  • EP3057493B1 patent drawingFigure 1
  • EP3057493B1 patent drawingFigure 2
  • EP3057493B1 patent drawingFigure 3

AI summary

A method and a system for assessing a condition in a subject. An example of a condition is a Major Depressive Disorder (MDD). The method comprises measuring at least one speech-related variable in a subject; extracting a channel-delay correlation structure of the at least one speech-related variable; and generating an assessment of a condition of the subject, based on the correlation structure of the at least one speech-related variable.