Acoustic activity detection apparatus and method

• US 9,502,028 B2
• Filed: 10/13/2014
• Issued: 11/22/2016
• Est. Priority Date: 10/18/2013
• Status: Active Grant

• Current Assignee: Knowles Electronics Llc (Knowles Corporation)
• Original Assignee: Knowles Electronics Llc (Knowles Corporation)
• Inventors: Nandy, Dibyendu, Li, Yang, Thomsen, Henrik, Furst, Claus
• Primary Examiner(s): Augustin, Marcellus
• Application Number:
  US14/512,877
• Publication Number:
  US 20150112689A1
• Time in Patent Office: 771 Days
• Field of Search: 704/270.1, 704/275, 704/273, 704/246, 704/233, 704/214
• US Class Current: 1/1
• CPC Class Codes: G06F 1/32 Means for saving power
• CPC Class Codes: G10L 15/20 Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech G10L21/02 takes precedence
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 19/002 Dynamic bit allocation for perceptual audio coders G10L19/032
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 25/84 for discriminating voice from noise

Abstract

Streaming audio is received. The streaming audio includes a frame having plurality of samples. An energy estimate is obtained for the plurality of samples. The energy estimate is compared to at least one threshold. In addition, a band pass estimate of the signal is determined. An energy estimate is obtained for the band-passed plurality of samples. The two energy estimates are compared to at least one threshold each. Based upon the comparison operation, a determination is made as to whether speech is detected.

Claims

1. An apparatus configured to distinguish speech activity from background noise, the apparatus comprising:

an analog circuit that converts sound energy into an analog electrical signal;

a conversion circuit coupled to the analog circuit that converts the analog signal into a digital signal;

a digital circuit coupled to the conversion circuit, the digital circuit including an acoustic activity detection (AAD) module, the AAD module configured to receive the digital signal, the digital signal comprising a sequence of frames, each frame having a plurality of samples, the AAD module configured to obtain an energy estimate for the plurality of samples of a frame and compare the energy estimate to at least one threshold, and the AAD module configured to determine whether speech or noise is detected based on the comparison, and when speech is detected to trigger transmission of an interrupt;

wherein the conversion circuit comprises a sigma-delta modulator that is configured to convert the analog signal into a single bit stream pulse density modulated (PDM) format;

wherein the digital circuit comprises a decimator module that converts the single bit stream pulse density modulated (PDM) format into a pulse code modulated (PCM) format;

wherein the pulse code modulated (PCM) audio from the decimator module is stored in a buffer while the AAD module determines whether speech or noise is detected.

2. The apparatus of claim 1, wherein the AAD module is further configured to determine whether a speech hangover has occurred.

3. The apparatus of claim 1, wherein the AAD module is further configured to update the threshold as a non-linear function of a noise threshold.

4. A microphone apparatus comprising:

a sensor having an output with an electrical signal produced in response to acoustic energy detected by the sensor;

a converter having an input coupled to the output of the sensor, the converter having an output with a digital signal obtained from the electrical signal;

a buffer coupled to the output of the converter, data based on the digital signal buffered in the buffer;

a voice activity detector coupled to the output of the converter, the voice activity detector distinguishing speech-like activity from non-speech based on a comparison of energy estimates of samples of data based on the digital signal to a threshold while the data is buffered,

the threshold determined at least in part by noise statistics that are independent of noise type;

an external-device interface coupled to the buffer, wherein a wake-up signal and data delayed by the buffer are provided to the external-device interface after the voice activity detector determines the presence of speech-like activity in the frame.

5. The apparatus of claim 4,

the energy estimate includes a first energy estimate for a first frequency band and a second energy estimate for a second frequency band, the first frequency band different than the second frequency band,

the voice activity detector distinguishing speech-like activity from non-speech based on a comparison of the first energy estimate and the second energy estimate to the threshold.

6. The apparatus claim 5, the first frequency band is a full bandwidth and the second frequency band is a relatively high frequency pass band, wherein the second frequency band captures sibilant and other fricative characteristics of speech.

7. The apparatus of claim 4,

a PCM format converter coupled to the output of the converter, the PCM format converter converting a pulse density modulated (PDM) format digital signal on the output of the converter to a pulse code modulated (PCM) format digital data,

the buffer coupled to the output of the PCM format converter, PCM format digital data buffered in the buffer, and

a PDM format converter interconnecting the buffer and the external-device interface, the PDM format converter converting PCM format digital data from the buffer to PDM format digital data, wherein PDM format digital data is provided to the external-device interface after determining the presence of speech-like activity.

8. The apparatus of claim 4,

the digital signal comprising a sequence of frames, each frame having a plurality of samples,

wherein data is no longer provided to the external-device interface after the voice activity detector determines the presence of non-speech in a frame.

9. The apparatus of claim 4,

the microphone apparatus continuing to provide data to the external-device interface for a duration after the voice activity detector determines the presence of non-speech before discontinuing providing data to the external-device interface,

the duration based on a comparison of an energy estimate for samples of data, in which the presence of non-speech was determined, with a non-linear combination of a noise threshold and a speech threshold.

10. The apparatus of claim 4, the threshold having a first update rate when non-speech is determined to be present and the threshold having a second update rate when speech is determined to be present, the second update rate less than the first update rate.

11. The apparatus of claim 4, the voice activity detector updating the threshold when non-speech is determined to be present and not updating the threshold when speech is determined to be present.

12. The apparatus of claim 4, providing the wake-up signal to the external-device interface by providing the data delayed by the buffer to the external-device interface, wherein the data delayed by the buffer is interpreted as the wake-up signal.

13. The apparatus of claim 4, the buffer is a circular buffer.

14. A method in a microphone apparatus having an acoustic sensor, a converter, a buffer, a voice activity detector, and an external-device interface, the method comprising:

generating an electrical signal in response to an acoustic input at the sensor;

converting the electrical signal to a digital signal using the converter;

distinguishing speech-like activity from non-speech by comparing an energy estimate for samples of data based on the digital signal to a threshold using the voice activity detector, the threshold determined at least in part by noise statistics that are independent of noise type;

buffering data based on the digital signal in the buffer while distinguishing speech-like activity from non-speech; and

providing a wake-up signal and data delayed by the buffer to the external-device interface after determining the presence of speech-like activity.

15. The method of claim 14,

the energy estimate includes a first energy estimate and a second energy estimate,

obtaining the first energy estimate for a first frequency band and obtaining the second energy estimate for a second frequency band, the first frequency band different than the second frequency band,

distinguishing speech-like activity from non-speech by comparing the first energy estimate and the second energy estimate to the at least one threshold using the voice activity detector.

16. The method claim 15, obtaining the first energy estimate for a full bandwidth and obtaining the second energy estimate for a relatively high frequency pass band, wherein the second frequency band captures sibilant and other fricative characteristics of speech.

17. The method of claim 14, discontinuing providing data to the external-device interface after determining a presence of non-speech while providing data to the external-device interface.

18. The method of claim 17, after determining the presence of non-speech while providing data to the external-device interface, continuing to provide data to the external-device interface for a duration that is based on a comparison of an energy estimate for samples of data, in which the presence of non-speech was determined, with a non-linear combination of a noise threshold and a speech threshold.

19. The method of claim 14,

converting the electrical signal to pulse density modulated (PDM) format digital data and converting the PDM format digital data to pulse code modulated (PCM) format digital data,

buffering the PCM format digital data while distinguishing speech-like activity from non-speech, and

converting the PCM format digital data to PDM format digital data before providing the data delayed by the buffer to the external-device interface after determining the presence of speech-like activity.

20. The method of claim 14, updating the threshold at a first rate when non-speech is determined to be present and updating the threshold at a second rate when speech is determined to be present, the second rate less than the first rate.

21. The method of claim 14, updating the threshold when non-speech is determined to be present and not updating the threshold when speech is determined to be present.

22. The method of claim 14, providing the wake-up signal to the external-device interface by providing the data delayed by the buffer to the external-device interface, wherein the data delayed by the buffer is interpreted as the wake-up signal.