Vehicle Speech Recognition Noise Filtering via Acoustic Model Adaptation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Voice-activated vehicle systems face challenges in accurately recognizing speech commands due to distortion from non-speech related sounds, such as engine noise, external noises, and audio system sounds, which reduces the clarity and decipherability of voice commands.

Innovation Solution

The system utilizes sound-related vehicle information, like engine RPM, HVAC settings, and external noise levels, to generate an interference profile record, which is used to modify the speech recognition process by applying filters and adapting acoustic models to enhance speech recognition and audio output.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If speech recognition is performed in vehicle environments, then voice commands can be processed, but non-speech related sounds (engine noise, external noises, audio system sounds) distort speech commands and reduce recognition accuracy

Engineering Contradiction:
Improvespeech recognition accuracyVSAvoidnon-speech related sounds
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent segments the audio signal into speech components and non-speech components (noise) using spectral subtraction and adaptive filtering. The speech recognition system processes only the separated speech signal, effectively isolating it from interfering vehicle noises, engine sounds, and external audio system sounds.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary noise estimation and cancellation module between the microphone input and speech recognition system. This intermediary component estimates background noise characteristics and subtracts them from the input signal, serving as a mediator that protects the speech recognition system from harmful noise while preserving speech integrity.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If traditional speech recognition systems are used in vehicles, then basic voice commands can be recognized, but the clarity and decipherability of voice commands are reduced by overpowering non-speech sounds

Engineering Contradiction:
Improvevoice command recognition reliabilityVSAvoidspeech signal clarity
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent implements feedback mechanisms where the noise estimation module continuously monitors the audio environment and adjusts its noise profile based on recent signal characteristics. This feedback loop allows the system to adapt to changing noise conditions (engine acceleration, HVAC changes, radio volume adjustments) and maintain reliable speech recognition throughout the vehicle operation cycle.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent performs preliminary noise estimation and spectral subtraction before the speech recognition process begins. By pre-processing the audio signal to remove estimated noise components in advance, the system ensures that the speech recognition module receives a cleaner signal, improving reliability before recognition attempts are made.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS9263040B2Method and system for using sound related vehicle information to enhance speech recognition
Publication Date: 2016.02.16 GM GLOBAL TECHNOLOGY OPERATIONS LLC
  • US9263040B2 patent drawing
  • US9263040B2 patent drawing
  • US9263040B2 patent drawing

AI summary

An audio signal may be received, in a processor associated with a vehicle. Sound related vehicle information representing one or more sounds may be received by the processor. The sound related vehicle information may or may not include an audio signal. A speech recognition process or system may be modified based on the sound related vehicle information.