Multi-Microphone Speaker Extraction for Speech Ambiguity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing audio systems and devices face challenges in effectively separating the speech of the user from interfering speech and noise due to speaker ambiguity and permutation issues, particularly in environments with multiple speakers.

Innovation Solution

An audio device equipped with a machine-learning model, such as a neural network, processes microphone input signals to extract near and far speaker components, allowing for improved separation of user speech from interfering speech and noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If single channel speech separation methods are used to suppress interfering speech, then interfering speech suppression is improved, but speaker ambiguity problem occurs

Engineering Contradiction:

Improveinterfering speech suppressionVSAvoidspeaker identification accuracy

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The patent transitions from single-channel to multi-channel microphone input, adding spatial dimension information. Multiple microphones capture speech from different spatial positions, enabling the system to distinguish between near speaker and far speaker based on spatial characteristics rather than relying solely on audio signal processing, thereby resolving speaker ambiguity while maintaining interfering speech suppression.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent introduces an extraction model as an intermediary component that processes multi-channel microphone inputs. This model acts as a mediator between the raw audio signals and the final speech separation output, extracting relevant features and resolving speaker identity ambiguity before producing the separated speech streams.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If classical speech separation methods are used, then processing complexity is reduced, but speaker permutation problem occurs

Engineering Contradiction:

Improveprocessing complexityVSAvoidspeaker separation accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary spatial filtering and beamforming operations to the multi-channel microphone inputs before feeding them to the extraction model. This preliminary action pre-processes the signals to enhance spatial separation characteristics, making the subsequent speech separation task easier and more accurate while avoiding the speaker permutation problem.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional mechanical signal processing methods with a machine learning-based extraction model. This substitution enables the system to learn complex spatial and spectral patterns automatically, achieving accurate speaker separation without relying on simplified classical methods that suffer from permutation ambiguity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12567434B2Audio system, audio device, and method for speaker extraction

Publication Date: 2026.03.03 GN HEARING AS

AI summary

A method for speech extraction in an audio device is disclosed. The method comprises obtaining a microphone input signal from one or more microphones including a first microphone. The method comprises applying an extraction model to the microphone input signal for provision of an output. The method comprises extracting a near speaker component in the microphone input signal according to the output of the extraction model being a machine-learning model for provision of a speaker output. The method comprises outputting the speaker output.