Environmental noise detection for dialog systems

• US 9,818,404 B2
• Filed: 12/22/2015
• Issued: 11/14/2017
• Est. Priority Date: 12/22/2015
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Firby, Robert Jim, Graff, Peter
• Primary Examiner(s): Godbold, Douglas
• Application Number:
  US14/979,189
• Publication Number:
  US 20170178627A1
• Time in Patent Office: 693 Days
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/221 Announcement of recognition results
• CPC Class Codes: G10L 2015/226 using non-speech characteristics
• CPC Class Codes: G10L 25/60 for measuring the quality of voice signals
• CPC Class Codes: G10L 25/72 for transmitting results of analysis
• CPC Class Codes: G10L 25/84 for discriminating voice from noise

Abstract

Embodiments are directed to receiving a speech signal representative of audible speech, processing the speech signal to interpret the speech signal by a dialog system implemented at least partially in hardware, determining, by the dialog system, that the speech signal cannot be correctly interpreted, receiving a noise signal representative of audible background noise, identifying a noise level from the noise signal, determining, by the dialog system, that the noise level is too high for the speech signal to be correctly interpreted, and providing, by the dialog system, a message indicating that the noise level is too high for the speech signal to be correctly interpreted.

Claims

1. A device comprising:

an audio input to receive audible speech from a user of the device and to receive audible background noise;

a sound signal processor to determine a noise level from the received audible background noise; and

a dialog system to;

receive a recognized speech signal;

determine that the recognized speech input is indecipherable;

determine that the noise level is greater than a threshold noise level; and

output a message indicating that the audible background noise is causing the audible speech to be incomprehensible;

wherein the dialog system is configured to;

store noise signals received over time;

correlate noise signals received over time and failed attempts to interpret received speech signals; and

determine the threshold noise level based on the correlation between noise signals received over time and the failed attempts to interpret received speech signals; and

use the threhsold noise level to determine that the speech signal incomprehensible.

2. The device of claim 1, further comprising an automatic speech recognition module implemented at least partially in hardware to receive a speech signal from the speech input, recognize the speech signal, and convert the recognized speech signal into a text format.

3. The device of claim 2, wherein automatic speech recognition module comprises the sound signal processor and the automatic speech recognition module is further configured to receive a noise signal from the audio input and determine a noise level of the audible background noise.

4. The device of claim 1, wherein the determined noise level comprises a signal-to-noise ratio.

5. The device of claim 1, further comprising a memory implemented at least partially in hardware to store a threshold noise level, and wherein

the dialog system is configured to:

compare a received noise level to the threshold noise level stored in memory;

determine that the received noise level exceeds the threshold noise level; and

output a message to the user that the background noise is too high.

6. The device of claim 1, further comprising a memory for storing historic noise data, wherein the dialog system is configured to:

dynamically determine that a received noise level is causing the audible speech to be incomprehensible.

7. The device of claim 1, wherein the dialog system is configured to:

determine that the audible speech signal is not understandable;

request a noise level representing an audible background noise level;

determine that the noise level causes the audible speech signal to be not understandable; and

output a message to the user indicating that the background noise level is too high to understand the audible speech signal.

8. The device of claim 1, wherein the dialog system is configured to:

determine that the audible speech signal is not understandable;

determine that the noise level is lower than a threshold value; and

output a request to the user to repeat the audible speech.

9. A method comprising:

receiving a speech signal representative of audible speech;

processing the speech signal to interpret the speech signal by a dialog system implemented at least partially in hardware;

determining, by the dialog system, that the speech signal cannot be correctly interpreted;

receiving a noise signal representative of audible background noise;

identifying a noise level from the noise signal;

determining, by the dialog system, that the noise level is too high for the speech signal to be correctly interpreted; and

providing, by the dialog system, a message indicating that the noise level is too high for the speech signal to be correctly interpreted;

the method further comprising;

storing noise signals received over time;

correlating noise signals received over time and failed attempts to interpret received speech signals; and

determining a threshold noise level based on the correlation between noise signals received over time and the failed attempts to interpret received speech signals; and

using the threhsold noise level to determine that the speech signal cannot be correctly interpreted.

10. The method of claim 9, wherein receiving the speech signal and receiving the noise signal occur at the same time.

11. The method of claim 9, further comprising requesting the noise signal based on determining that the speech signal cannot be correctly interpreted.

12. The method of claim 9, further comprises converting the speech signal into recognized text and providing the speech signal as recognized text to the dialog system.

13. The method of claim 9, wherein determining, by the dialog system, that the noise level is too high for the speech signal to be correctly interpreted comprises determining that the noise level is greater than, or greater than or equal to, a threshold value.

14. The method of claim 9, further comprising:

determining that the speech signal cannot be correctly interpreted;

determining that the noise level is less than a threshold value; and

requesting that the user repeat a speech signal representative of the audible speech.

15. A system comprising:

a microphone to receive an audible speech signal and audible background noise;

a signal processor to determine a noise level of the audible background noise;

an automatic speech recognition system implemented at least partially in hardware to convert the audible speech signal to a text format; and

a dialog system implemented at least partially in hardware to;

attempt to interpret the text format of the audible speech signal;

determine that the noise level of the audible background noise causes the text format of the audible speech to be uninterpretable; and

provide a message that indicates that the audible speech cannot be interpreted because the audible background noise too high;

wherein the dialog system is configured to;

store noise signals received over time;

correlate noise signals received over time and failed attempts to interpret received speech signals; and

determine a threshold noise level based on the correlation between noise signals received over time and the failed attempts to interpret received speech signals; and

use the threhsold noise level to determine that the speech signal is uninterpretable.

16. The system of claim 15, wherein the automatic speech recognition module is further configured to receive a speech signal from the microphone, recognize the speech signal, and convert the recognized speech signal into a text format.

17. The system of claim 15, wherein the automatic speech recognition module comprises the sound signal processor and the automatic speech recognition module is further configured to receive a noise signal from the microphone and determine a noise level of the audible background noise.

18. The system of claim 15, wherein the determined noise level comprises a signal-to-noise ratio.

19. The system of claim 15, further comprising a memory implemented at least partially in hardware to store a threshold noise level, and wherein

the dialog system is configured to:

compare a received noise level to the threshold noise level stored in memory;

determine that the received noise level exceeds the threshold noise level; and

output a message to the user that the background noise is too high.

20. The system of claim 15, further comprising a memory for storing historic noise data, wherein the dialog system is configured to:

dynamically determine that a received noise level is causing the audible speech to be incomprehensible.

21. The system of claim 15, wherein the dialog system is configured to:

determine that the audible speech signal is not understandable;

determine that the noise level causes the audible speech signal to be not understandable; and

output a message to the user indicating that the background noise level is too high to understand the audible speech signal.

22. The system of claim 15, wherein the dialog system is configured to:

determine that the audible speech signal is not understandable;

determine that the noise level is lower than a threshold value; and

output a message to the user requesting the user to repeat the audible speech signal.