Hearing device comprising a noise reduction system

Patent metadata

Field	Value
Publication number	US-11245993-B2
Application number	US-202016785167-A
Country	US
Kind code	B2
Filing date	Feb 7, 2020
Priority date	Feb 8, 2019
Publication date	Feb 8, 2022
Grant date	Feb 8, 2022

Abstract

A hearing device, e.g. a hearing aid, is configured to be worn by a user at or in an ear or to be fully or partially implanted in the head at an ear of the user. The hearing device comprises a) an input unit for providing at least one electric input signal in a time frequency representation k, m, where k and m are frequency and time indices, respectively, and k represents a frequency channel, the at least one electric input signal being representative of sound and comprising target signal components and noise components; and b) a signal processor comprising b1) an SNR estimator for providing a target signal-to-noise ratio estimate for said at least one electric input signal in said time frequency representation; and b2) an SNR-to-gain converter for converting said target signal-to-noise ratio estimate to respective gain values in said time frequency representation. The signal processor comprises a neural network, wherein the weights of the neural network have been trained with a plurality of training signals. A method of operating a hearing aid is further disclosed. The invention may e.g. be used in audio devices, such as hearing aids, headsets, mobile telephones, etc., operating in noisy acoustic environments.

First claim

The invention claimed is:

1. A hearing device, configured to be worn by a user at or in an ear or to be fully or partially implanted in the head at an ear of the user, the hearing device comprising
an input unit for providing at least one electric input signal in a time frequency representation k, m, where k and m are frequency and time indices, respectively, and k represents a frequency channel, the at least one electric input signal being representative of sound and comprising target signal components and noise components; and
a signal processor comprising
an SNR estimator unit for providing a target signal-to-noise ratio (SNR) estimate for said at least one electric input signal in said time frequency representation;
an SNR-to-gain converter for converting said target signal-to-noise ratio estimates to respective gain values in said time frequency representation, wherein

said signal processor comprises a neural network, wherein the weights of the neural network have been trained with a plurality of training signals.

2. A hearing device according to claim 1 wherein said SNR estimator and/or said SNR-to-gain converter comprises said neural network.

3. A hearing device according to claim 1 wherein said SNR estimator unit comprises first and second SNR estimators.

4. A hearing device according to claim 3 wherein said first and second SNR estimators are sequentially coupled, so that the output of the first SNR estimator is used by the second SNR estimator to provide an improved SNR estimate.

5. A hearing device according to claim 4 wherein the output of said second SNR estimator is used as input to said SNR-to-gain converter.

6. A hearing device according to claim 3 wherein the outputs of said first and second SNR estimators are used in parallel as inputs to said SNR-to-gain converter.

7. A hearing device according to claim 3 wherein said first SNR estimator is configured to provide said first target signal-to-noise ratio estimate independently in each frequency channel.

8. A hearing device according to claim 3 wherein said second SNR estimator comprises said neural network, and wherein the weights of the neural network have been trained with the plurality of training signals.

9. A hearing device according to claim 1 wherein said SNR-to-gain converter comprises said neural network, wherein the weights of the neural network have been trained with the plurality of training signals.

10. A hearing device according to claim 1 wherein said SNR-to-gain converter implements a non-linear function G(k,m), k=1, . . . , K, where G is gain, and wherein gain G(k,m) in the k th frequency-channel depends on said target signal-to-noise ratio estimates of one or more further, such as all K, frequency-channels at time index m, and optionally on previous values of said estimates, and wherein said non-linear function is implemented by said neural network.

11. A hearing device according to claim 1 wherein the neural network is optimized towards only partly attenuating the noise components of the at least one electric input signal(s).

12. A hearing device according to claim 1 wherein the SNR estimator unit and/or the SNR-to-gain converter is configured to receive additional inputs from one or more sensors or detectors.

13. A hearing device according to claim 12 wherein said one or more sensor or detectors provide one or more of
a (single or multichannel) voice activity flag,
a (single or multichannel) own voice activity flag,
a different SNR estimate,
an onset flag,
estimated Direction of Arrival (DoA) information, and
a camera based input capturing lip-reading or throat movement information.

14. A hearing device according to claim 13 wherein a different SNR estimate is based on signal modulation, or spatial properties utilizing at least two microphone signals, or binaural SNR estimates.

15. A hearing device according to claim 13 wherein the onset flag is provided by an onset or transient detector derived directly from a time domain input signal.

16. A hearing device according to claim 1 wherein said SNR-to-gain converter is configured to provide a maximum amount of noise reduction.

17. A hearing device according to claim 16 configured to provide that said maximum amount of noise reduction is dependent on the type of noise.

18. A hearing device according to claim 1 being constituted by or comprising a hearing aid, a headset, an earphone, an ear protection device or a combination thereof.

19. A hearing device according to claim 1 wherein said neural network comprises a convolutional neural network or a recurrent neural network.

20. A method of operating a hearing device configured to be worn by a user at or in an ear or to be fully or partially implanted in the head at an ear of the user, the method comprising
providing at least one electric input signal in a time frequency representation k, m, where k and m are frequency and time indices, respectively, and k represents a frequency channel, the at least one electric input signal being representative of sound and comprising target signal components and noise components;
providing a target signal-to-noise ratio estimate for said at least one electric input signal in said time frequency representation;
converting said target signal-to-noise ratio estimate to respective gain values in said time frequency representation; and
providing said target signal-to-noise ratio estimate, and/or said respective gain values, using a neural network, wherein the weights of the neural network have been trained with a plurality of training signals.

21. A binaural hearing system comprising first and second hearing devices, each of the first and second hearing devices configured to be worn by a user at or in an ear or to be fully or partially implanted in the head at an ear of the user, each of the first and second hearing devices comprising:
an input unit for providing at least one electric input signal in a time frequency representation k, m, where k and m are frequency and time indices, respectively, and k represents a frequency channel, the at least one electric input signal being representative of sound and comprising target signal components and noise components; and
a signal processor comprising
an SNR estimator for providing a target signal-to-noise ratio (SNR) estimate for said at least one electric input signal in said time frequency representation;
an SNR-to-gain converter for converting said target signal-to-noise ratio estimates to respective gain values in said time frequency representation,

wherein said signal processor of each of the first and second hearing devices comprises a neural network, wherein the weights of the neural network have been trained with a plurality of training signals, and
wherein the first and second hearing devices are adapted to establish a wireless link between them and to exchange data between them, wherein said data include said target SNR-estimates, and wherein the SNR-to-gain converter of the first and second hearing devices are configured to include said target SNR estimates of the respective opposite hearing device in the estimation of respective first and second gain values in said time frequency representation.

22. A hearing aid configured to be worn by a user at or in an ear or to be fully or partially implanted in the head at an ear of the user, the hearing aid comprising
an input unit for providing at least one electric input signal in a time frequency representation k, m, where k and

Assignees

• Oticon As

Inventors

• Andersen Asger Heidemann
• Jensen Jesper
• Pedersen Michael Syskind
• Rohde Nels Hede
• Olsen Anders Brødløs
• Kristensen Michael Smed
• Bentsen Thomas
• Bertelsen Andreas Thelander

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0495

  Quantised networks; Sparse networks; Compressed networks · CPC title

• H04R2225/43

  Signal processing in hearing aids to enhance the speech intelligibility · CPC title