Speech Recognition Correction for Impaired Speech

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional speech recognition systems fail to effectively recognize impaired speech, leading to difficulties for individuals with speech disorders in using voice input devices, as these systems often misinterpret or fail to recognize user commands.

Innovation Solution

A processor-implemented method and system that processes speech signals by identifying deviation models for various speech disorders, determining the extent and location of deviations, and applying speech modifications to generate healthy speech, utilizing feature vectors and deep learning techniques like Deep Auto Encoders and Hybrid Deep Neural Networks for improved recognition.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional speech recognition systems are used, then normal speech can be recognized accurately, but impaired speech recognition accuracy deteriorates

Engineering Contradiction:
Improvespeech recognition accuracyVSAvoidcapability to handle impaired speech
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary analysis of the speech signal to detect deviations from normal speech patterns before recognition. By identifying characteristics of impaired speech in advance and comparing against deviation models, the system prepares correction strategies beforehand, enabling accurate recognition of impaired speech without compromising normal speech recognition performance

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system introduces deviation models as an intermediary between the speech signal and recognition process. These models serve as reference standards that mediate the comparison between actual speech and expected patterns, enabling the system to handle both normal and impaired speech by detecting and correcting deviations from these intermediate reference points

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If speech modification techniques are applied to correct impaired speech, then recognition accuracy improves, but system complexity increases

Engineering Contradiction:
Improveimpaired speech recognition accuracyVSAvoidsystem architecture complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the speech processing task into distinct modules: deviation detection, model comparison, deviation level determination, and speech modification. Each module handles a specific aspect of the correction process, making the overall complex system manageable and maintainable while achieving high recognition accuracy for impaired speech

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system dynamically adapts its processing based on the detected deviation level. By determining the extent of impairment and applying appropriate modification strategies tailored to each case, the system achieves high accuracy without requiring all possible correction methods to be always active, thus managing complexity through conditional, adaptive processing

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3618061B1Method and system for improving recognition of disordered speech
Publication Date: 2022.04.27 TATA CONSULTANCY SERVICES LTD
  • EP3618061B1 patent drawingFigure 1
  • EP3618061B1 patent drawingFigure 2
  • EP3618061B1 patent drawingFigure 3

AI summary

Speech recognition is a technique that enables recognition and translation of spoken languages (speech data) into text, using a computer. This allows the users of a system having speech recognition to provide voice commands for various purposes. However, existing systems for (automatic) speech recognition fails or finds it difficult to interpret impaired speech. Disclosed herein is a method and a system for identifying extent of deviation in speech utterances of a user from a normal level, caused due to such impairments, and for making appropriate modifications to generate utterances pertaining to healthy speech. This data may be fed as input to the speech recognition systems, as those systems can interpret the corrected data.