Methods and apparatus for unsupervised wakeup with time-correlated acoustic events
Methods and apparatus for unsupervised wakeup of a device including receiving a first acoustic event at a first time and a second acoustic event at a second time, wherein the first and second acoustic events have scores above a first threshold identifying the first and second acoustic events as wakeup candidates for a wakeup phrase for an unsupervised wakeup of a device. It can be determined that the first acoustic event score is below a second threshold, which is higher than the first threshold and whether a difference between the first and second times is within a range to check for correlation in time between the first and second acoustic events. Occurrence of a wakeup event can be determined based upon the first and second times.
1. A method, comprising:
receiving a first acoustic event at a first time and a second acoustic event at a second time from a microphone of a device, wherein the first and second acoustic events have scores above a first threshold identifying the first and second acoustic events as wakeup candidates for a wakeup phrase for an unsupervised wakeup of the device;
storing the first and second acoustic events and the first and second times in memory on the device;
determining, using a computer processor on the device, that the first acoustic event score is below a second threshold, which is higher than the first threshold; and
determining, using the computer processor, whether a difference between the first and second times is within a range to check for correlation in time between the first and second acoustic events; and
determining that a wakeup event has occurred based upon the first and second times and performing wakeup of the device.
2. The method according to claim 1 , further including determining whether the first and second acoustic events are acoustically similar and determining that the wakeup event has occurred based upon the first and second times and the determination of acoustic similarity.
3. The method according to claim 2 , further including adapting an acoustic model on the first and/or second acoustic event based on the acoustic similarity determination of the first and second acoustic events.
4. The method according to claim 2 , further including adapting an acoustic model on a lower scoring one of the first and second acoustic events.
5. The method according to claim 1 , wherein an upper limit of the range is less than or equal to about fifteen seconds.
6. The method according to claim 1 , wherein a lower limit of the range is greater than or equal to about three seconds.
7. The method according to claim 1 , further including using a main processor and a lower power processor to provide the unsupervised wakeup.
8. The method according to claim 7 , wherein the lower power processor identifies the first wakeup candidate.
9. The method according to claim 1 , further including receiving a textual input of the wakeup phrase.
10. The method according to claim 1 , further including receiving at least a third acoustic event above the first threshold and below the second threshold within the range to determine that the wakeup event has occurred.
11. The method according to claim 10 , further including determining acoustic similarity of the at least a third acoustic event with at least one of the first and second acoustic events to determine that the wakeup event has occurred.
12. An article, comprising:
a non-transitory computer readable medium having stored instructions that enable a machine to:
receive a first acoustic event at a first time and a second acoustic event at a second time from a microphone of a device, wherein the first and second acoustic events have scores above a first threshold identifying the first and second acoustic events as wakeup candidates for a wakeup phrase for an unsupervised wakeup of the device;
store the first and second acoustic events and the first and second times in memory on the device;
determine, using a computer processor on the device, that the first acoustic event score is below a second threshold, which is higher than the first threshold; and
determine, using the computer processor, whether a difference between the first and second times is within a range to check for correlation in time between the first and second acoustic events; and
determine that a wakeup event has occurred based upon the first and second times and perform a wakeup of the device.
13. The article according to claim 12 , further including instructions to determine whether the first and second acoustic events are acoustically similar and determining that the wakeup event has occurred based upon the first and second times and the determination of acoustic similarity.
14. The article according to claim 13 , further including instructions to adapt an acoustic model on the first and/or second acoustic event based on the acoustic similarity determination of the first and second acoustic events.
15. The article according to claim 12 , wherein an upper limit of the range is less than or equal to about fifteen seconds.
16. The article according to claim 12 , wherein a lower limit of the range is greater than or equal to about three seconds.
17. The article according to claim 12 , further including instructions for using a main processor and a lower power processor to provide the unsupervised wakeup.
18. A system, comprising:
a processor and a memory configured to:
receive a first acoustic event at a first time and a second acoustic event at a second time from a microphone of a device, wherein the first and second acoustic events have scores above a first threshold identifying the first and second acoustic events as wakeup candidates for a wakeup phrase for an unsupervised wakeup of the device;
store the first and second acoustic events and the first and second times in memory on the device;
determine, using the processor, that the first acoustic event score is below a second threshold, which is higher than the first threshold; and
determine, using the processor, whether a difference between the first and second times is within a range to check for correlation in time between the first and second acoustic events; and
determine that a wakeup event has occurred based upon the first and second times and perform a wakeup of the device.
19. The system according to claim 18 , wherein the processor and memory are further configured to include determining whether the first and second acoustic events are acoustically similar and determining that the wakeup event has occurred based upon the first and second times and the determination of acoustic similarity.
20. The system according to claim 19 , wherein the processor and memory are further configured to adapt an acoustic model on the first and/or second acoustic event based on the acoustic similarity determination of the first and second acoustic events.