US 12,003,930 B2Granted

Neural network driven acoustic feedback detection in audio system

OwnerThomas Damgaard, Pauline O'Callaghaninventors

Life of this patent

Estimating the term…

Loading patent…

Abstract

A method and device for detecting acoustic feedback events with an artificial neural network in an in-ear earbud audio system that allows, by user interaction, playback of a recorded and processed signal from an environment-recording microphone by an in-ear speaker that faces or is at least acoustically coupled with the ear canal such that sound played by the speaker enters the ear canal. The audio system employs an acoustic seal that acoustically separates the speaker from the microphone, but due to external factors, the acoustical separation may not be adequate, thereby forming acoustic feedback paths. The neural network facilitates a binary classification of the time-wise segmented microphone signal, which is used to stop playback by the in-ear speaker if a feedback event is detected to protect the hearing of the user. Detection of a feedback event triggers an audible or wireless notification to be delivered to the user.

Claims18

Dependent claims

An audio device comprising:

a. an environment microphone configured and arranged to convert an input sound signal to an electrical input signal,
b. a speaker configured and arranged to convert an electrical output signal to an output sound,
c. an acoustic insulator that is configured to acoustically separate said environment microphone from said speaker when the acoustic insulator is fitted into an ear canal of a user,
d. an audio system comprising:
a neural network configured to receive and analyse at least samples of said electrical input signal from the environment microphone and to compute an acoustic feedback event classification of said at least samples of said electrical input signal,
a processing unit configured to receive the electrical input signal from the environment microphone and to forward an electrical output signal based on the electrical input signal to the speaker, the electrical output signal being generated by the processing unit, wherein the processing unit is configured to forward the electrical output signal based on said acoustic feedback event classification,
wherein said acoustic feedback event classification is a binary classification corresponding to “feedback detected” or “feedback not detected”.

4 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/720,347Filed 14 Apr 2022 · us provisional application
Priority: 15 Apr 2021
Published: US 2022/0337949 A120 Oct 2022
Granted: 4 Jun 2024
Term ends: Not estimatedThe source records don’t support an estimate.

Classification12 codes

H04R 3/02
Circuits for transducers: for preventing acoustic reaction , i.e. acoustic oscillatory feedback
G06N 3/0499
Architecture, e.g. interconnection topology: feedforward networks
G06N 3/08
Neural networks: learning methods
G06N 3/09
Learning methods: supervised learning

Family

Application63/175,127Us provisional application · filed 15 Apr 2021
US 2022/0337949 A1Pre-grant publication · published 20 Oct 2022
US 12,003,930 B2This patent

From the US filing history. Foreign family members aren’t listed.

Timeline

Citations

Open elsewhere

Building an agent or pulling many patents? Use the API or MCP server instead of scraping pages. Get an API key