Granted Patent B2

US 12,323,768 · App. 18/089,093 · Granted Jun 3, 2025

Inventors: Paul Wagner (Portland, OR); Sebastian Kroedel (Erlenbach, CH); Claudio Santelli (Staefa, CH); Markus Hofbauer (Hombrechtikon, CH); Manon Barbier (Soultzmatt, FR)

Assignee: Sonova AG

H04R25/507H04R25/606

Abstract

Claims (31)

1. A method for audio signal processing on a hearing system, the hearing system having an audio input unit, a processing unit for audio signal processing, an audio output unit, and a neural network for audio signal processing, the method comprising:

receiving an input audio signal by means of the audio input unit,

audio signal processing the input audio signal into an output audio signal at least partially by means of the processing unit, the audio signal processing including

executing audio signal processing by the neural network for performing a step of the audio signal processing,

determining a confidence parameter resembling a reliability of the audio signal processing by the neural network, and

steering the audio signal processing in dependence of the confidence parameter,

outputting the output audio signal by means of the audio output unit.

2. The method according to claim 1 , wherein the confidence parameter determines whether a network output of the neural network is used in further processing steps of the audio signal processing.

3. The method according to any claim 1 , wherein the confidence parameter is determined by the neural network.

4. The method according to claim 1 , wherein the neural network performs a classification task, in particular performs audio scene classification.

5. The method according to claim 1 , wherein the neural network performs regression-based acoustic processing to obtain network output audio data.

6. The method according to claim 5 , wherein the neural network outputs the confidence parameter additionally to the network output audio data.

7. The method according to claim 5 , wherein the regression-based acoustic processing by the neural network is modified in dependence of the confidence parameter.

8. The method according to claim 1 , wherein the output audio signal comprises a mixture of a primary audio signal, which is obtained using audio signal processing by the neural network, and a secondary audio signal, which is obtained without using audio signal processing by the neural network, and wherein a mixing ratio of primary audio signal and secondary audio signal is determined based on the confidence parameter.

9. The method according to claim 1 , wherein the input audio signal is processed by a secondary processing routine to obtain a secondary audio signal.

10. The method according to claim 1 , wherein an audio signal to be processed by the neural network is based on a transformation of the input audio signal received over a time interval, wherein a length of the time interval depends on the confidence parameter.

11. The method according to claim 1 , wherein a sample of the input audio signal is recorded in situations, when the confidence parameter indicates a low reliability of the audio signal processing by the neural network.

12. A hearing system, comprising:

an audio input unit for receiving an input audio signal,

a processing unit for audio signal processing the input audio signal into an output audio signal,

an audio output unit for outputting the output audio signal, and

a neural network for audio signal processing which, when executed, contributes to the audio signal processing,

wherein the hearing system is configured to

determine a confidence parameter resembling a reliability of the audio signal processing by the neural network, and

steer the audio signal processing by the neural network based on the confidence parameter.

13. The hearing system according to claim 12 , further comprising a mixing unit for mixing a primary audio signal, which is obtained using audio signal processing by the neural network, with a secondary audio signal, which is obtained without using audio signal processing by the neural network.

14. The hearing system according to claim 12 , wherein the hearing system is implemented by one or more of a hearing aid, a cochlear implant, or a bone-anchored hearing aid.

15. A neural network for audio signal processing, wherein the neural network is configured to:

perform a regression-based acoustic processing of an audio signal to obtain network output audio data,

determine a confidence parameter resembling a reliability of regression-based acoustic processing, and

output the confidence parameter and/or to modify the regression-based acoustic processing in dependence of the confidence parameter.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (19)