Non-parametric voice activity detection

• US 6,023,674 A
• Filed: 01/23/1998
• Issued: 02/08/2000
• Est. Priority Date: 01/23/1998
• Status: Expired due to Term

• Current Assignee: IDTP Holdings Incorporated (InterDigital, Inc.)
• Original Assignee: Telefonaktiebolaget LM Ericsson
• Inventors: Mekuria, Fisseha
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US09/012,518
• Time in Patent Office: 746 Days
• Field of Search: 704/233, 704/207, 704/206, 704/200, 704/201, 704/226, 704/222, 704/216, 704/217, 704/218, 704/210
• US Class Current: 704/233
• CPC Class Codes: G10L 2025/783 based on threshold decision
• 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

Speech or voice activity in an audio signal is defected without using a speech coder. Pitch period information and signal energy information are extracted from the audio signal, and a decision regarding the presence or absence of voice is made based on that information.

Claims

1. A method of detecting a speech component signal in an audio signal, comprising:

extracting from the audio signal information about a pitch period of the speech component signal without using information obtained from a speech coder;

extracting signal energy information from the audio signal; and

deciding whether the speech component signal is present in the audio signal based on the information about the pitch period and the signal energy information.

2. The method of claim 1, wherein said first-mentioned extracting step includes processing the audio signal to produce a signal having peaks that are separated by the pitch period of the speech component signal.

3. The method of claim 2, wherein said processing step includes applying low pass and non-linear filtering to the audio signal to remove from the audio signal information that is not indicative of the pitch period.

4. The method of claim 1, wherein said last-mentioned extracting step includes suppressing first signal values of the audio signal that are less than a predetermined threshold value, and multiplying second signal values of the audio signal that exceed the predetermined threshold value by respective multiplier values that vary as a function of the second signal values.

5. The method of claim 4, wherein said last-mentioned extracting step includes squaring the magnitudes of the multiplied second signal values.

6. The method of claim 5, wherein said last-mentioned extracting step includes dividing the squared magnitude values into groups, and summing the squared magnitude values of each group.

7. The method of claim 6, wherein said dividing step includes selecting members for each group based on the information about the pitch period.

8. The method of claim 4, wherein said last-mentioned extracting step includes low pass filtering the audio signal before performing said suppressing and multiplying steps, and performing said suppressing and multiplying steps on the low pass filtered audio signal.

9. The method of claim 4, wherein the multiplier values vary linearly with the second signal values.

10. An apparatus for detecting a speech component signal in an audio signal, comprising:

a pitch period detector which extracts from the audio signal information about a pitch period of the speech component signal without using information obtained from a speech coder;

a signal energy detector which extracts signal energy information from the audio signal; and

a decision section that is connected to said pitch period detector and to said signal energy detector and which decides whether the speech component signal is present in the audio signal based on the information about the pitch period and the signal energy information.

11. The apparatus of claim 10, wherein said pitch period detector includes a signal processing section that processes the audio signal to produce a signal having peaks that are separated by the pitch period of the speech component signal.

12. The apparatus of claim 11, wherein said signal processing section includes a low pass filter section and a non-linear filter section which remove from the audio signal information that is not indicative of the pitch period.

13. The apparatus of claim 10, wherein said signal energy detector includes a multiplier that (1) suppresses first signal values of the audio signal that are less than a predetermined threshold value and (2) multiplies second signal values of the audio signal that exceed the threshold value by respective multiplier values that vary as a function of the second signal values.

14. The apparatus of claim 13, wherein said signal energy detector includes a magnitude squaring section that squares the magnitudes of said multiplied second signal values, said magnitude squaring section connected to said multiplier.

15. The apparatus of claim 14, wherein said signal energy detector includes a summing section that operates on a group of said squared magnitude values and sums said group of squared magnitude values, said summing section connected to said squaring section.

16. The apparatus of claim 15, wherein said summing section is also connected to said pitch period detector to receive said information about said pitch period for defining said group of squared magnitude values.

17. The apparatus of claim 13, including a low pass filter which is selectively connectable to an input of said multiplier for low pass filtering said audio signal and passing a low pass filtered audio signal to said multiplier.

18. The apparatus of claim 13, wherein the multiplier values vary linearly with the second signal values.

19. A mobile telecommunications terminal, comprising:

a microphone for receiving an input audio signal;

a digitizer coupled to said microphone for digitizing the audio signal; and

an apparatus coupled to said digitizer for detecting a speech component signal in the digitized audio signal, said apparatus including a pitch period detector which extracts from the digitized audio signal information about a pitch period of the speech component signal without using information obtained from a speech coder, a signal energy detector which extracts signal energy information from the digitized audio signal, and a decision section that is connected to said pitch period detector and to said signal energy detector and which decides whether the speech component signal is present in the digitized audio signal based on the information about the pitch period and the signal energy information.