Method and apparatus for a low power voice trigger device
View Patent ↗Aspects of the present disclosure involve a method for a voice trigger device that can be used to interrupt an externally connected system. The current disclosure also presents the architecture for the voice trigger device used for searching and matching an audio signature with a reference signature. In one embodiment a reverse matching mechanism is performed. In another embodiment, the reverse search and match operation is performed using an exponential normalization technique.
1. A method, comprising:
configuring a voice trigger circuit to:
receive a training sequence and store a representation of the training sequence in a training buffer in reverse of order received;
detect receipt of an audio signal that is sampled into blocks;
determine a plurality of energy values of the sampled audio signal blocks;
perform energy binning of the plurality of energy values to determine whether speech is present in the sampled audio signal blocks;
determine that speech is present in the sampled audio signal blocks received;
store a representation of the sampled audio signal block in a trigger buffer;
match the representation of the sampled audio signal blocks stored in the trigger buffer starting with the representation of most recently received sampled audio signal block and proceeding to oldest received sampled audio signal block, to the representation of the training sequence stored in the training buffer starting with the most recently received and proceeding to oldest received; and
enable a wake up pin in the voice trigger circuit upon matching the representation to the training sequence.
2. The method of claim 1 , wherein the audio signal is received by a mixed signal component in the voice trigger device.
3. The method of claim 1 , wherein the sampled audio signal blocks are a trigger sequence.
4. The method of claim 1 , wherein the plurality of energy values are determined using a frequency domain translated block of samples.
5. The method of claim 1 , wherein the energy binning reduces a number of bins used to process each of the blocks of samples received.
6. The method of claim 1 , wherein Mel-Frequency Cepstrum Coefficients (MFCCs) are determined using at least the energy binning.
7. The method of claim 6 , wherein the MFCCs are exponentially normalized.
8. The method of claim 6 , wherein the MFCCs are used at least in part to represent the training sequence.
9. The method of claim 6 , wherein a distance is determined for the MFCCs, and wherein speech is present if the distance is less than a pre-determined threshold.
10. A system for voice trigger device wake up comprising:
an I/O processing unit, the I/O processing unit configured to:
receive an audio signal;
a core processing unit, the core processing unit configured to:
detect receipt of the audio signal that is sampled into blocks;
an overlap-add processing unit, the overlap-add processing unit configured to:
determine a plurality of energy values of the sampled audio signal blocks; the core processing unit further configured to:
perform energy binning of the plurality of energy values to determine whether speech is present in the sampled audio signal blocks;
determine that speech is present in the sampled audio signal blocks received;
match a trigger buffer to a training sequence stored in a training buffer in reverse of order received, wherein a representation of the sampled audio signal blocks is stored in the trigger buffer and wherein the match is performed starting with the representation of a most recently received audio signal matched to the most recently received of the training sequence that is stored in the training buffer; and
enable a wake up pin in a voice trigger device upon matching the trigger buffer to the training buffer.
11. The system of claim 10 , wherein the overlap-add processing unit is a weighted overlap-add (WOLA) processor, and wherein the WOLA processor is configured to apply a window to the blocks of samples received.
12. The method of claim 11 , wherein the WOLA processor translates the blocks of samples received into a frequency domain.
13. The system of claim 10 , wherein the core processing unit is configured to reduce a number of bins used to process the blocks of samples received by performing the energy binning.
14. The system of claim 10 , wherein Mel-Frequency Cepstrum Coefficients (MFCCs) are determined using at least the energy binning.
15. The system of claim 14 , wherein the MFCCs are exponentially normalized.
16. The system of claim 14 , wherein the MFCCs are used at least in part to represent the training sequence.
17. The system of claim 14 , wherein the core processing unit is configured to determine a distance for the MFCCs, and wherein the core processing unit is further configured to determine that speech is present if the distance is less than a pre-determined threshold.
18. A non-transitory computer-readable data storage medium comprising instructions that, when executed by at least one processor of a device, cause the device to perform operations comprising:
detecting receipt of an audio signal that is sampled into blocks;
determining a plurality of energy values of the sampled audio signal blocks;
performing energy binning of the plurality of energy values to determine whether speech is present in the sampled audio signal blocks;
determining that speech is present in the sampled audio signal blocks received;
storing a representation of the sampled audio signal block in a trigger buffer;
matching the representation of the sampled audio signal blocks stored in the trigger buffer to a training sequence stored in a training buffer in reverse of order received wherein the matching proceeds from matching the representation of a most recent sampled audio signal to a most recently sampled of the training sequence; and
enabling a wake up pin in a voice trigger device upon matching the representation to the training sequence.