Neural Network Audio Filter for Speech Impairment Correction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Individuals with speech impairments face challenges in effectively communicating through software-based audio communication platforms, as existing technologies lack practical solutions to correct distorted or repetitive speech impairments in real-time.

Innovation Solution

A system utilizing a neural network audio filter, comprising a convolutional neural network (CNN) and recurrent neural network (RNN) with connectionist temporal classification (CTC), that identifies and removes impaired sections from speech signals in real-time, enabling the formation of unimpaired messages.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional audio communication platforms are used, then communication between individuals is enabled, but speech impaired messages cannot be effectively corrected or improved

Engineering Contradiction:

Improvecommunication effectivenessVSAvoidaccommodation for speech impairments

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

A neural network audio filter is introduced as an intermediary component between the speech impaired message and the communication platform. The filter processes the audio signal in real-time, identifying and correcting impairments such as stuttering and repetitions, thereby enabling effective communication without requiring changes to the user's speech patterns or the communication platform itself

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system dynamically changes audio signal parameters by applying neural network processing to the audio stream. The filter modifies temporal and spectral characteristics of the speech signal to remove impairments while preserving the core message, effectively transforming the audio parameters to achieve clear communication

Inventive Principle:

Principle #35Parameter changes

2Reliability

If real-time speech correction is implemented, then speech impaired messages are corrected, but system complexity increases

Engineering Contradiction:

Improvespeech correction accuracyVSAvoidneural network filter complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Traditional mechanical or rule-based speech correction systems are replaced with a neural network-based audio filter. This substitution enables more accurate and adaptive speech impairment correction while maintaining real-time processing capability, as the neural network learns optimal correction strategies from training data rather than relying on fixed rules

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If speech impairments are corrected in real-time, then communication clarity is improved, but processing time increases

Engineering Contradiction:

Improvemessage clarityVSAvoidprocessing delay

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The neural network filter is pre-trained on extensive speech data to learn common speech impairment patterns and their corrections. This preliminary training enables the filter to perform real-time correction without requiring complex computations during actual use, thereby minimizing processing delay while maintaining high message clarity

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11682413B2Method and system to modify speech impaired messages utilizing neural network audio filters

Publication Date: 2023.06.20 LENOVO SWITZERLAND INTERNATIONAL GMBH

AI summary

A computer implemented method, system and computer program product are provided that implement a neural network (NN) audio filter. The method, system and computer program product obtain an electronic audio signal comprising a speech impaired message and apply the audio signal to the NN audio filter to modify the speech impaired message to form an unimpaired message. The method, system and computer program product output the unimpaired message.