System and method for processor wake-up based on sensor data

• US 9,349,386 B2
• Filed: 03/07/2013
• Issued: 05/24/2016
• Est. Priority Date: 03/07/2013
• Status: Active Grant

• Current Assignee: Analog Devices International Unlimited Company (Analog Devices, Inc.)
• Original Assignee: Analog Devices, Inc.
• Inventors: Adams, Robert, Mortensen, Maikael
• Primary Examiner(s): Zhu, Richard
• Application Number:
  US13/788,062
• Publication Number:
  US 20140257821A1
• Time in Patent Office: 1,174 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 1/3206 Monitoring of events, devices or parameters that trigger a change in power modality
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 2025/783 based on threshold decision
• CPC Class Codes: G10L 2025/786 Adaptive threshold
• CPC Class Codes: G10L 21/06 Transformation of speech into a non-audible representation, e.g. speech visualisation or speech processing for tactile aids G10L15/26 takes precedence
• CPC Class Codes: G10L 25/48 specially adapted for particular use
• 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
• CPC Class Codes: G10L 25/84 for discriminating voice from noise
• CPC Class Codes: G10L 25/87 Detection of discrete points within a voice signal
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence

Abstract

A system for processor wake-up based on sensor data includes an audio buffer, an envelope buffer, and a processor. The audio buffer is configured to store a first data from a sensor. The first data is generated according to a first sampling rate. The envelope buffer is configured to store a second data, which is derived from the first data according to a second sampling rate, which is less than the first sampling rate. The processor is configured to wake up periodically from an idle state and read the second data from the envelope buffer. If the second data indicates an activity, the processor is configured to read the first data from the audio buffer. If the second data does not indicate an activity, the processor is configured to return to the idle state.

Claims

1. A method executed by a voice activity detector (VAD) comprising a processor, wherein the method comprises:

generating a first data from a sensor according to a first sampling rate;

deriving a second data from a normalized envelope of a subset of the first data according to a second sampling rate that is less than the first sampling rate;

comparing the second data with a first negative peak value;

detecting an activity if a difference between the second data and the first peak value exceeds a predetermined threshold;

comparing the second data with a second negative peak value;

incrementing the second negative peak value by a fixed amount to obtain the first negative peak value if the second data is not less than the second negative peak value during a predetermined number of clock cycles; and

setting the first negative peak value to the second negative peak value if the second data is less than the second negative peak value.

2. The method of claim 1, further comprising storing the first data in an audio buffer of the VAD.

3. The method of claim 2, further comprising storing the second data in an envelope buffer of the VAD, wherein the envelope buffer is shorter than the audio buffer.

4. The method of claim 2, further comprising:

if the activity is detected, reading the first data from the audio buffer; and

if the activity is not detected, returning the processor to an idle state.

5. The method of claim 1, further comprising sending periodically, by a timer of the VAD, an interrupt to the processor, wherein the processor wakes up on receiving the interrupt and reads the second data.

6. The method of claim 1, further comprising

generating, by a bandpass filter of the VAD, the subset of the first data;

detecting, by an envelope detector of the VAD, an envelope of the subset; and

normalizing the envelope.

7. The method of claim 6, wherein the normalizing the envelope comprises taking a ratio of energies represented by the first data.

8. The method of claim 6, wherein the normalizing the envelope comprises a log normalization of the envelope.

9. The method of claim 1, wherein the first data is generated from audio signals.

10. The method of claim 9, wherein frequencies beyond a normal human voice range are filtered out of the audio signals to generate the first subset.

11. The method of claim 1, wherein the activity comprises an onset of a voice signal.

12. The method of claim 1, wherein the detecting the envelope is based on a parameter of the subset selected from a group consisting of:

root mean square (RMS), average, peak and quasi-peak.

13. The method of claim 1, further comprising:

determining that the activity occurred between two consecutive wake-up times of the processor;

executing a keyword matching algorithm on the first data; and

if a matching keyword is found, performing another task based on the matching keyword.

14. A voice activity detector (VAD) apparatus, comprising:

an audio buffer that stores a first data from a sensor according to a first sampling rate;

a bandpass filter that generates a subset of the first data;

an envelope detector that detects an envelope of the subset and normalizes the envelope;

an envelope buffer that stores a second data derived from the normalized envelope according to a second sampling rate that is less than the first sampling rate; and

a processor that reads the second data from the envelope buffer and detects an activity if the second data is greater by a predetermined threshold than a first negative peak value, wherein the processor further compares the second data with a second negative peak value, wherein if the second data is not less than the second negative peak value during a predetermined number of clock cycles, the processor increments the second negative peak value by a fixed amount to obtain the first negative peak value, and if the second data is less than the second negative peak value, the processor sets the first negative peak value to the second negative peak value.

15. The VAD apparatus of Claim 14, wherein the processor returns to an idle state if the second data does not indicate the activity.

16. The VAD apparatus of Claim 14, further comprising a direct memory access (DMA) controller that facilitates reading the audio buffer and the envelope buffer.

17. The VAD apparatus of Claim 14, further comprising an analog-to-digital converter (ADC) that generates the first data in a digital format according to the first sampling rate from an analog signal at the sensor.

18. The VAD apparatus of Claim 14, wherein the envelope buffer is shorter than the audio buffer.

19. The VAD apparatus of Claim 14, further comprising a decimator, wherein the decimator derives the second data by downsampling the normalized envelope.

20. The VAD apparatus of Claim 14, wherein the bandpass filter is configured to filter the first data and feed the filtered data to the envelope detector.