US 10,872,599 B1 · Wakeword training

Abstract

A device monitors audio data for a predetermined and/or user-defined wakeword. The device detects an error in detecting the wakeword in the audio data, such as a false-positive detection of the wakeword or a false-negative detection of the wakeword. Upon detecting the error, the device updates a model trained to detect the wakeword to create an updated trained model; the updated trained model reduces or eliminates further errors in detecting the wakeword. Data corresponding to the updated trained model may be collected by a server from a plurality of devices and used to create an updated trained model aggregating the data; this updated trained model may be sent to some or all of the devices.

Claims24

Dependent claims

A computer-implemented method comprising:

generating first audio data from first audio, the first audio detected by a first device during a first time period;
determining, using a model, a first probability that the first audio data includes a representation of at least a portion of a word;
determining that the first probability corresponds to at least one of a false positive detection of the at least the portion of the word or a false negative detection of the at least the portion of the word;
and
determining, in response to the first probability corresponding to the at least one of the false positive detection or the false negative detection, error data corresponding to a difference between the representation of at least the portion of the word and a stored representation of at least the portion of the word.

2 more elements

A system comprising:

at least one processor;
and
at least one memory including instructions that, when executed by the at least one processor, cause the system to:
generate first audio data from first audio, the first audio detected by a first device during a first time period;
determine, using a model, a first probability that the first audio data includes a representation of at least a portion of a word;
determine that the first probability corresponds to at least one of a false positive detection of the at least the portion of the word or a false negative detection of the at least the portion of the word;
and
determine, in response to the first probability corresponding to the at least one of the false positive detection or the false negative detection, error data corresponding to a difference between the representation of at least the portion of the word and a stored representation of at least the portion of the word.

5 more elements

Dependencies are inferred from the claim text.

Wakeword training

Life of this patent

Wakeword training

Life of this patent

Abstract

Claims24