Multiple range dynamic level control

• US 9,171,552 B1
• Filed: 01/17/2013
• Issued: 10/27/2015
• Est. Priority Date: 01/17/2013
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Yang, Jun
• Primary Examiner(s): PULLIAS, JESSE SCOTT
• Application Number:
  US13/744,134
• Time in Patent Office: 1,013 Days
• Field of Search: 704/200, 704/200.1, 704/201, 704/225, 348/462, 348/736, 348/738, 381/104, 381/107
• US Class Current: 1/1
• CPC Class Codes: G10L 21/0316 by changing the amplitude
• CPC Class Codes: G10L 25/84 for discriminating voice from noise

Abstract

An audio-based system may perform dynamic level adjustment by detecting voice activity in an input signal and evaluating voice levels during periods of voice activity. The current voice level is compared to a plurality of thresholds to determine a corresponding gain strategy, and the input signal is scaled in accordance with this gain strategy. Further adjustment to the signal is performed to reduce output clipping that might otherwise be produced.

Claims

1. A computing device, comprising:

a processor;

one or more microphones configured to generate an input audio signal;

one or more speakers; and

memory, accessible by the processor and storing instructions that are executable by the processor to perform acts in multiple repetitions, the acts of each repetition comprising;

detecting voice presence in the input audio signal;

determining a voice level associated with the voice presence in the input audio signal;

comparing the voice level to at least one of a plurality of threshold amplitudes, each threshold amplitude of the plurality of threshold amplitudes corresponding to one of multiple level ranges;

identifying one of the multiple level ranges to which the voice level corresponds based at least in part on the comparing;

selecting an audio gain based at least in part on the identified one of the multiple level ranges;

smoothing the selected audio gain over time;

scaling the input audio signal by the selected and smoothed audio gain to produce an intermediate audio signal; and

attenuating the intermediate audio signal to reduce clipping, wherein the attenuating produces an output audio signal for output by the one or more speakers.

2. The computing device of claim 1, wherein detecting the voice presence comprises performing noise activity detection (NAD) with respect to the input audio signal.

3. The computing device of claim 1, wherein detecting the voice presence comprises estimating a signal envelope and a noise floor of the input audio signal.

4. The computing device of claim 1, wherein:

the smoothing is performed by a first order low-pass filter having a first time constant that limits the rate of change of the selected and smoothed audio gain over time; and

the attenuating is applied to peaks of the intermediate audio signal with a compressor having a second time constant that is shorter than the first time constant.

5. The computing device of claim 1 wherein:

the input audio signal comprises a left input audio signal and a right input audio signal corresponding to left and right stereo channels, respectively; and

determining the voice level comprises determining a maximum of;

(i) a voice level of the left input audio signal, and (ii) a voice level of the right input audio signal.

6. A method of dynamically controlling an audio level, comprising:

specifying a plurality of thresholds to define multiple level ranges and corresponding gain strategies;

detecting voice presence in one or more audio signals, the one or more audio signals including the voice presence and other noise;

determining a voice level associated with the voice presence in the one or more audio signals;

comparing the voice level to the plurality of thresholds to identify one of the multiple level ranges to which the determined voice level corresponds; and

selecting an audio gain based at least in part on the identified one of the multiple level ranges.

7. The method of claim 6, further comprising applying the selected audio gain to the one or more audio signals to create one or more output audio signals.

8. The method of claim 6, further comprising smoothing the selected audio gain over time.

9. The method of claim 6, further comprising:

applying the selected audio gain to the one or more audio signals to create one or more intermediate audio signals; and

attenuating peaks of the one or more intermediate audio signals to reduce clipping.

10. The method of claim 6, further comprising:

smoothing the selected audio gain over time using a first time constant;

applying the selected and smoothed audio gain to produce one or more intermediate audio signals; and

attenuating peaks of the one or more intermediate audio signals to reduce clipping, wherein the attenuating is performed using a second time constant that is shorter than the first time constant.

11. The method of claim 6, wherein detecting the voice presence comprises performing noise activity detection (NAD) with respect to the one or more audio signals.

12. The method of claim 6, wherein detecting the voice presence comprises estimating a signal envelope and a noise floor of the one or more audio signals.

13. One or more non-transitory computer-readable media storing computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform acts comprising:

detecting voice presence in one or more audio signals, the one or more audio signals including the voice presence and other noise;

determining a voice level associated with the voice presence in the one or more audio signals;

specifying a plurality of thresholds to define multiple level ranges and corresponding gain strategies;

comparing the voice level to the plurality of thresholds to identify one of multiple level ranges to which the voice level corresponds;

selecting an audio gain based at least in part on the identified one of the multiple level ranges; and

applying the selected audio gain to the one or more audio signals.

14. The one or more non-transitory computer-readable media of claim 13, further comprising smoothing the selected audio gain over time.

15. The one or more non-transitory computer-readable media of claim 13, wherein applying the selected audio gain produces one or more intermediate audio signals, the acts further comprising attenuating peaks of the one or more intermediate audio signals to reduce clipping.

16. The one or more non-transitory computer-readable media of claim 13, wherein applying the selected audio gain produces one or more intermediate audio signals, the acts further comprising:

smoothing the selected audio gain over time using a first time constant; and

attenuating peaks of the one or more intermediate audio signals to reduce clipping, wherein the attenuating is performed using a second time constant that is shorter than the first time constant.

17. The one or more non-transitory computer-readable media of claim 13, wherein detecting the voice presence comprises performing noise activity detection (NAD) with respect to the one or more audio signals.

18. The one or more non-transitory computer readable media of claim 13, wherein detecting the voice presence comprises estimating a signal envelope and a noise floor of the one or more audio signals.

19. The one or more non-transitory computer-readable media of claim 13, wherein the one or more audio signals comprise left and right audio signals corresponding to left and right stereo channels, respectively.

20. The one or more non-transitory computer-readable media of claim 13, wherein the other noise includes stationary noise.