Method for operating a hearing device and hearing device

Patent metadata

Field	Value
Publication number	US-10609494-B2
Application number	US-201816102988-A
Country	US
Kind code	B2
Filing date	Aug 14, 2018
Priority date	Aug 14, 2017
Publication date	Mar 31, 2020
Grant date	Mar 31, 2020

Abstract

A method operates a hearing device for a user. In the method, electromyography is performed, in which a muscle activity of an auricular muscle of the user is measured by an electrode array. The muscle activity is continuously measured to detect a complex activity profile of the auricular muscle, in which the user's intention is coded. The electrode array generates a sensor signal that is classified by means of a classifier, and in this classification process, the muscle activity is decoded and the underlying intention is determined by examining whether the sensor signal has a previously known feature vector. The previously known feature vector is assigned to an operating mode of the hearing device, which is set when the sensor signal has the previously known feature vector.

First claim

The invention claimed is:

1. A method for operating a hearing device of a user, which comprises the steps of:
carrying out an electromyography step in which muscle activity of an auricular muscle of the user is measured using an electrode array;
continuously measuring the muscle activity to detect a complex activity profile of the auricular muscle in which an underlying intention of the user is coded;
generating, via the electrode array, a sensor signal that is classified by means of a classifier in which the muscle activity is decoded and the underlying intention is determined by examining whether the sensor signal has a previously known feature vector; and
assigning the previously known feature vector to an operating mode of the hearing device that is set if the sensor signal has the previously known feature vector.

2. The method according to claim 1 , which further comprises measuring the muscle activity over a measurement period of at least 100 ms and that the previously known feature vector is searched for in the sensor signal over an entire measurement period.

3. The method according to claim 1 , which further comprises:
configuring the hearing device for directional hearing; and
adjusting a directivity of the hearing device on a basis of the muscle activity, and a sound source in a specific direction is emphasized.

4. The method according to claim 3 , wherein the hearing device is a binaural hearing device and has two individual devices to be worn on different sides of a head of the user, and the sensor signal is generated on each of the different sides by means of the electrode array and the muscular activity of the auricular muscle is measured.

5. The method according to claim 4 , wherein the muscle activities on both of the different sides are first evaluated separately by classifying sensor signals independently of one another, and then, if results agree, the hearing device is set.

6. The method according to claim 4 , wherein muscle activities on both of the different sides are first evaluated separately by classifying the sensor signals independently of one another, and if results differ, the individual devices are synchronized by adjusting one of the individual devices against its own result and with reference to a result of the other of the individual devices.

7. The method according to claim 4 , wherein the underlying intention is a desired hearing direction, which is determined by examining which of the two sensor signals has a higher degree of organization or a higher energy, and the directivity is adjusted toward the hearing direction.

8. The method according to claim 1 , wherein an own-voice situation, in which the user is speaking, is recognized based on the muscle activity, and the operating mode is an own-voice reduction in which a natural voice component is reduced in an output signal of the hearing device.

9. The method according to claim 1 , wherein a facial expression of the user is recognized on a basis of the muscle activity, and is used as a control command for setting a specific operating mode that is assigned to the control command.

10. The method according to claim 1 , wherein faulty positioning of the hearing device is recognized based on the muscle activity.

11. The method according to claim 1 , which further comprises setting an amplification scheme, a compression scheme or a noise suppression of the hearing device, based on the muscle activity.

12. The method according to claim 1 , wherein the previously known feature vector is stored in an external database.

13. The method according to claim 1 , which further comprises subjecting the sensor signal to feature extraction by means of a feature extractor, in which a new feature vector is generated which is stored as an additional previously known feature vector.

14. The method according to claim 13 which further comprises generating the new feature vector in a supervised machine learning process, wherein:
a particular operating mode is assigned to a particular situation;
the particular situation is generated as a current situation;
the feature extractor independently searches for a number of unknown features in the sensor signal;
the unknown features are combined to form the new feature vector; and
the new feature vector is assigned to the particular operating mode.

15. The method according to claim 14 , which further comprises carrying out the supervised machine learning process during normal operation of the hearing device, with the feature extractor continuously examining whether the new feature vector is contained in the sensor signal.

16. The method according to claim 14 , wherein:
by means of an additional sensor it is recognized whether the current situation is a previously known situation already associated with a previously known operating mode; and
if the current situation corresponds to the previously known situation, the new feature vector is assigned to the previously known operating mode.

17. The method according to claim 13 , which further comprises integrating the feature extractor and the classifier together into a signal processor, and the feature extraction and classification are treated as a single optimization problem.

18. The method according to claim 14 , which further comprises integrating the previously known feature vector into a neural network, by means of the neural network it is examined whether the sensor signal has the previously known feature vector, by determining a feature vector of the sensor signal and using it as an input signal for the neural network.

19. The method according to claim 18 , which further comprises recognizing by means of an additional sensor whether the current situation is a previously known situation already associated with a previously known operating mode, and if an integrated feature vector coincides with the feature vector of the sensor signal, the feature vector is integrated into the neural network which is assigned to the previously known operating mode.

20. The method according to claim 1 , which further comprises dividing the sensor signal into a plurality of frequency bands by a filter bank, with each frequency band forming a number of features.

21. The method according to claim 20 , wherein the filter bank is a paraunitary and parameterized filter bank.

22. The method according to claim 20 , wherein the filter bank has a lower limit frequency of 4 Hz and an upper limit frequency of 1 kHz, so that the muscle activity is examined in a frequency range from 4 Hz to 1 kHz.

23. The method according to claim 1 , wherein the electrode array has at most five electrodes.

24. The method according to claim 1 , wherein the electrode array has a plurality of electrodes, one of the electrodes is a reference electrode, by means of which a reference signal is measured away from the auricular muscle of the user, and in that the sensor signal is prepared by means of the reference signal before the sensor signal is fed to the classifier.

25. A hearing device, comprising:
a control unit for operating the hearing device, said control unit programmed to:
carry out an electromyography step in which muscle activity of an auricular muscle of a user is measured using an electrode array;
continuously measure the muscle activity to detect a complex activity profile of the auricular muscle in which an underlying intention of the user is coded;
generate, via the electrode array, a sensor

Assignees

• Sivantos Pte Ltd

Inventors

• Hannemann Ronny
• Strauss Daniel J
• Corona-Strauss Farah I

Classifications

• H04R25/507

  implemented by neural network or fuzzy logic · CPC title

• H04R25/405

  by combining a plurality of transducers · CPC title

• H04R25/305

  Self-monitoring or self-testing · CPC title

• H04R25/70

  Adaptation of deaf aid to hearing loss, e.g. initial electronic fitting · CPC title

• H04R2460/01

  Hearing devices using active noise cancellation · CPC title