Device and method for privacy-preserving vocal interaction

Patent metadata

Field	Value
Publication number	US-11069362-B2
Application number	US-201816213187-A
Country	US
Kind code	B2
Filing date	Dec 7, 2018
Priority date	Dec 7, 2017
Publication date	Jul 20, 2021
Grant date	Jul 20, 2021

Abstract

A home assistant device and a method for privacy-preserving vocal interactions. A microphone captures an audio signal corresponding to a vocal user query. The identity of the speaker is determined, and an obfuscated name is generated corresponding to the identified speaker. The audio signal is analysed to determine the intent of the user and a personalized answer is generated in combination with the obfuscated name. This answer is then de-obfuscated by reintroducing the speaker name. The de-obfuscated answer is rendered to the speaker.

First claim

The invention claimed is:

1. A device for performing privacy-preserving vocal interactions comprising:
a microphone configured to capture an audio signal representative of a vocal utterance relative to a query from a speaker;
a speaker identifier configured to determine the identity of a speaker from the captured audio signal;
a privacy enforcer configured to provide a first obfuscated speaker name corresponding to the identified speaker and store a list of correspondences between speaker names and obfuscated speaker names;
a communication interface configured to:
provide to an external device the captured audio signal and the first obfuscated speaker name;
receive from an external device an answer to the speaker query;

wherein the privacy enforcer is further configured to determine if the received answer contains the first obfuscated speaker name and in this case to replace, in the received answer, the first obfuscated speaker name by its corresponding name in the list.

2. The device of claim 1 further comprising providing the answer to the speaker.

3. The device of claim 1 wherein the received answer is in textual form.

4. The device of claim 3 further comprising a text-to-speech converter configured to transform the de-obfuscated answer from textual form into an audio signal.

5. The device of claim 1 wherein the received answer is in audio format.

6. The device of claim 1 wherein the privacy enforcer is further configured to obfuscate the captured audio signal by detecting a speaker name of the list and replacing it by an audio signal representative of the corresponding obfuscated speaker name.

7. The device of claim 1 wherein the privacy enforcer further comprises a setting to turn an incognito mode on or off wherein in the first case, the privacy enforcer obfuscates the speaker name before providing it and de-obfuscates the received answer when it contains an obfuscated speaker name, and in the second case, the privacy enforcer no more obfuscates the speaker name and no more determines if the received answer contains an obfuscated speaker name.

8. The device of claim 1 further comprising renewing the obfuscated speaker names of the list.

9. A method for performing privacy preserving vocal interactions comprising:
capturing an audio signal representative of a vocal utterance relative to a query from a speaker;
identifying a speaker from the captured audio signal;
generating a first obfuscated speaker name corresponding to the identified speaker and storing a list of correspondence between speaker names and obfuscated speaker names;
providing to an external device the captured audio signal and the first obfuscated speaker name;
obtaining from an external device an answer to the query; and
in response to a determination that the received answer contains the first obfuscated speaker name, replacing, in the received answer, the first obfuscated speaker name by its corresponding name in the list.

10. The method of claim 9 further comprising providing the answer to the speaker.

11. The method of claim 9 wherein the received answer is in textual form.

12. The method of claim 9 wherein the received answer is in audio form.

13. The method of claim 9 further comprising obfuscating the captured audio signal by detecting a speaker name of the list and replacing it by an audio signal representative of the corresponding obfuscated speaker name.

14. The method of claim 9 further comprising a setting to turn an incognito mode on or off wherein in the first case, obfuscating the speaker name before providing it and de-obfuscating the received answer when it contains an obfuscated speaker name, and in the second case, no more obfuscating the speaker name and no more determining if the received answer contains an obfuscated speaker name.

15. The method of claim 9 further comprising renewing the obfuscated speaker names of the list.

16. A non-transitory computer readable medium comprising program code instructions executable by a processor for implementing the method comprising:
capturing an audio signal representative of a vocal utterance relative to a query from a speaker;
identifying a speaker from the captured audio signal;
generating a first obfuscated speaker name corresponding to the identified speaker and storing a list of correspondence between speaker names and obfuscated speaker names;
providing to an external device the captured audio signal and the first obfuscated speaker name;
obtaining from an external device an answer to the query; and
in response to a determination that the received answer contains the first obfuscated speaker name, replacing, in the received answer, the first obfuscated speaker name by its corresponding name in the list.

17. The medium of claim 16 , wherein the method further comprises providing the answer to the speaker.

18. The medium of claim 16 wherein the received answer is in textual form.

19. The medium of claim 16 wherein the received answer is in audio form.

20. The medium of claim 16 , wherein the method further comprises obfuscating the captured audio signal by detecting a speaker name of the list and replacing it by an audio signal representative of the corresponding obfuscated speaker name.

Assignees

• Interdigital Ce Patent Holdings

Inventors

• Fontaine Patrick
• Neumann Christoph

Classifications

• G10L17/22 Primary

  Interactive procedures; Man-machine interfaces · CPC title

• G10L17/02 Primary

  Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction · CPC title

• G10L15/04

  Segmentation; Word boundary detection · CPC title

• G06F3/16

  Sound input; Sound output (speech processing G10L) · CPC title

• G10L13/08

  Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination · CPC title