Voice activity detector for speech signals in variable background noise

• US 5,649,055 A
• Filed: 09/29/1995
• Issued: 07/15/1997
• Est. Priority Date: 03/26/1993
• Status: Expired due to Term

• Current Assignee: Hughes Network Systems LLC (Echostar Corporation)
• Original Assignee: Hughes Electronics Corporation (Rtx Corporation)
• Inventors: Lamkin, Allan B., Morris, Adrian J., Kepley, W. Robert III, Gupta, Prabhat K., Jangi, Shrirang
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): COLLINS, ALPHONSO
• Application Number:
  US08/536,507
• Time in Patent Office: 655 Days
• Field of Search: 395/2, 395/2.17, 395/2.19, 395/2.23, 395/2.24, 395/2.35-2.37, 395/2.42, 395/2.55, 395/2.57, 395/2.62, 395/2.22, 381/42, 381/46
• US Class Current: 704/233
• CPC Class Codes: G10L 2025/786 Adaptive threshold
• CPC Class Codes: G10L 25/09 the extracted parameters being zero crossing rates
• CPC Class Codes: G10L 25/78 Detection of presence or absence of voice signals switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10

Abstract

A voice activity detector (VAD) which determines whether received voice signal samples contain speech by deriving parameters measuring short term time domain characteristics of the input signal, including the average signal level and the absolute value of any change in average signal level, and comparing the derived parameter values with corresponding thresholds, which are periodically monitored and updated to reflect changes in the level of background noise, thereby minimizing clipping and false alarms.

Claims

1. A method of detecting voice activity in received voice signal samples including background noise, comprising the steps of:

deriving voice signal parameters from the voice signal samples, wherein the voice signal parameters include an average signal level, calculated as a short-term average energy of the voice signal samples, and a slope, calculated as an absolute value of a change in the average signal level;

comparing the voice signal parameters with voice signal parameter thresholds and setting a Voice Activity Detection (VAD) flag according to the results of the comparisons;

updating the voice signal parameter thresholds at a first frequency to ensure rapid tracking of the background noise if the VAD flag is not set; and

updating the voice signal parameter thresholds at a second slower frequency for slower tracking of the background noise if the VAD flag is set.

2. The method of detecting voice activity as recited in claim 1, wherein the voice signal parameters further include a zero crossing count.

3. The method of detecting voice activity as recited in claim 2, wherein the zero crossing count is calculated over a sliding window.

4. The method of detecting voice activity as recited in claim 2, wherein the step of comparing the voice signal parameters with voice signal parameter thresholds further comprises the steps of:

comparing the average signal level with a high level threshold and setting the VAD flag if the average signal level is above the high level threshold;

but

if the average signal level is not above the high level threshold, then comparing the average signal level with a low level threshold and setting the VAD flag if the average signal level is above the low level threshold and either the slope is above a slope threshold or the zero crossing count is above a zero crossing count threshold.

5. The method of detecting voice activity as recited in claim 1, wherein:

the step of updating the voice signal parameter thresholds at the first frequency comprises updating in accordance with a first update time constant for controlling the first frequency; and

the step of updating the voice signal parameter thresholds at the second frequency comprises updating in accordance with a second update time constant for controlling the second frequency.

6. A voice activity detector for detecting voice activity in received voice signal samples including background noise, comprising:

a calculator for calculating voice signal parameters from the voice signal samples, the voice signal parameters including;

an average signal level, calculated as a short-term average energy of the voice signal samples; and

a slope, calculated as an absolute value of a change in the average signal level;

a comparator for comparing the voice signal parameters with voice signal parameter thresholds, wherein a Voice Activity Detection (VAD) flag is set based on the comparisons; and

an updater for updating the voice signal parameter thresholds at a first frequency to ensure rapid tracking of the background noise if the VAD flag is not set, and updating the voice signal parameter thresholds at a second slower frequency for slower tracking of the background noise if the VAD flag is set.

7. The voice activity detector of claim 6, wherein the voice signal parameters calculated by the calculator further include a zero crossing count.

8. The voice activity detector of claim 7, wherein the zero crossing count is calculated over a sliding window.

9. The voice activity detector of claim 7, wherein the comparator compares the average signal level with a high level threshold and sets the VAD flag if the average signal level is above the high level threshold;

but if the average signal level is not above the high level threshold, the comparator compares the average signal level with a low level threshold and sets the VAD flag if the average signal level is above the low level threshold and either the slope is above a slope threshold or the zero crossing count is above a zero crossing count threshold.

10. The voice activity detector of claim 6, wherein the updater updates the voice signal parameter thresholds at the first frequency in accordance with a first update time constant for controlling the first frequency, and updates the voice signal parameter thresholds at the second frequency in accordance with a second update time constant for controlling the second frequency.

11. A memory device storing instructions to be implemented by a data processor in a communications system, for detecting voice activity in received voice signal samples including background noise, the instructions comprising:

instructions for deriving voice signal parameters from the voice signal samples, wherein the voice signal parameters include an average signal level, calculated as a short-term average energy of the voice signal samples, and a slope, calculated as an absolute value of a change in the average signal level;

instructions for comparing the voice signal parameters with voice signal parameter thresholds and setting a Voice Activity Detection (VAD) flag according to the results of the comparisons;

instructions for updating the voice signal parameter thresholds at a first frequency to ensure rapid tracking of the background noise if the VAD flag is not set; and

instructions for updating the voice signal parameter thresholds at a second slower frequency for slower tracking of the background noise if the VAD flag is set.

12. The memory device of claim 11, wherein the voice signal parameters further include a zero crossing count.

13. The memory device of claim 12, wherein the zero crossing count is calculated over a sliding window.

14. The memory device of claim 12, wherein the instructions for comparing the voice signal parameters with voice signal parameter thresholds further comprises:

instructions for comparing the average signal level with a high level threshold and setting the VAD flag if the average signal level is above the high level threshold, but if the average signal level is not above the high level threshold, then comparing the average signal level with a low level threshold and setting the VAD flag if the average signal level is above the low level threshold and either the slope is above a slope threshold or the zero crossing count is above a zero crossing count threshold.

15. The memory device of claim 11, wherein the stored instructions further comprise:

instructions for updating the voice signal parameter thresholds at the first frequency in accordance with a first update time constant for controlling the first frequency; and

instructions for updating the voice signal parameter thresholds at the second frequency in accordance with a second update time constant for controlling the second frequency.

16. A voice activity detector for detecting voice activity in received voice signal samples comprising:

means for deriving voice signal parameters from the voice signal samples, including means for calculating an average signal level as a short-term average energy of the voice signal samples, and means for calculating a slope as an absolute value of a change in the average signal level;

means for comparing the voice signal parameters with voice signal parameter thresholds;

means for setting a Voice Activity Detection (VAD) flag according to the results of the comparisons;

means for updating the voice signal parameter thresholds at a first frequency to ensure rapid tracking of the background noise if the VAD flag is not set; and

means for updating the voice signal parameter thresholds at a second slower frequency for slower tracking of the background noise if the VAD flag is set.

17. The voice activity detector recited in claim 16, wherein the means for deriving voice signal parameters further includes means for calculating a zero crossing count.

18. The voice activity detector recited in claim 17, wherein the means for calculating the zero crossing count calculates the zero crossing count over a sliding window.

19. The voice activity detector recited in claim 17, wherein the means for comparing the voice signal parameters with voice signal parameter thresholds compares the average signal level with a high level threshold and sets the VAD flag if the average signal level is above the high level threshold;

but if the average signal level is not above the high level threshold, the means for comparing compares the average signal level with a low level threshold and sets the VAD flag if the average signal level is above the low level threshold and either the slope is above a slope threshold or the zero crossing count is above a zero crossing count threshold.

20. The voice activity detector recited in claim 16, wherein:

the means for updating the voice signal parameter thresholds at the first frequency updates in accordance with a first update time constant for controlling the first frequency; and

the means for updating the voice signal parameter thresholds at the second frequency updates in accordance with a second update time constant for controlling the second frequency.