Google Inc.

Alexander Houston Gruenstein

Diego Melendo Casado

Matthew Nirvan Sharifi

Petar Aleksic

Johan Schalkwyk

William J Byrne

Jakob Nicolaus Foerster

Kenneth Edward Mixter

Christopher Thaddeus Hughes

Terry Tai

Taral Pradeep Joglekar

Pedro Jose Moreno Mengibar

Ignacio Lopez Moreno

Ian C McGraw

Jakob Foerster

Douglas H Beeferman

Michiel A U Bacchiani

Pu-sen Chao

Aleksandar Kracun

Meltem Oktem

Vijayaditya Peddinti

Fnu Heryandi

Salil Rajadhyaksha

Quan Wang

Li Wan

Jason William Pelecanos

Aishanee Shah

Dave Harwath

Hakan Erdogan

Michael J LeBeau

Maria Carolina Parada San Martin

Nicholas Jitkoff

Guoguo Chen

Aleksandar Kacun

In one aspect, a method comprises accessing audio data generated by a computing device based on audio input from a user, the audio data encoding one or more user utterances. The method further comprises generating a first transcription of the utterances by performing speech recognition on the audio data using a first speech recognizer that employs a language model based on user-specific data. The method further comprises generating a second transcription of the utterances by performing speech recognition on the audio data using a second speech recognizer that employs a language model independent of user-specific data. The method further comprises determining that the second transcription of the utterances includes a term from a predefined set of one or more terms. The method further comprises, based on determining that the second transcription of the utterance includes the term, providing an output of the first transcription of the utterance.

Mixed language model speech recognition

A method for automatic hotword threshold tuning includes receiving, from a user device executing a first stage hotword detector configured to detect a hotword in streaming audio, audio data characterizing the detected hotword. The method includes processing, using a second stage hotword detector, the audio data to determine whether the hotword is detected by the second stage hotword detector. When the hotword is not detected, the method includes identifying a false acceptance instance at the first stage hotword detector indicating that the first stage hotword detector incorrectly detected the hotword. The method includes determining whether a false acceptance rate satisfies a false acceptance rate threshold based on a number of false acceptance instances within a false acceptance time period. When the false acceptance rate satisfies the false acceptance rate threshold, the method includes adjusting the hotword detection threshold of the first stage hotword detector.

Automatic hotword threshold tuning

A computing system receives requests from client devices to process voice queries that have been detected in local environments of the client devices. The system identifies that a value that is based on a number of requests to process voice queries received by the system during a specified time interval satisfies one or more criteria. In response, the system triggers analysis of at least some of the requests received during the specified time interval to trigger analysis of at least some received requests to determine a set of requests that each identify a common voice query. The system can generate an electronic fingerprint that indicates a distinctive model of the common voice query. The fingerprint can then be used to detect an illegitimate voice query identified in a request from a client device at a later time.

Detecting and suppressing voice queries

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for detecting hotwords using a server. One of the methods includes receiving an audio signal encoding one or more utterances including a first utterance; determining whether at least a portion of the first utterance satisfies a first threshold of being at least a portion of a key phrase; in response to determining that at least the portion of the first utterance satisfies the first threshold of being at least a portion of a key phrase, sending the audio signal to a server system that determines whether the first utterance satisfies a second threshold of being the key phrase, the second threshold being more restrictive than the first threshold; and receiving tagged text data representing the one or more utterances encoded in the audio signal when the server system determines that the first utterance satisfies the second threshold.

Server side hotwording

The various implementations described herein include methods and systems for determining device leadership among voice interface devices. In one aspect, a method is performed at a first electronic device of a plurality of electronic devices, each having microphones, a speaker, processors, and memory storing programs for execution by the processors. The first device detects a voice input. It determines a device state and a relevance of the voice input. It identifies a subset of electronic devices from the plurality to which the voice input is relevant. In accordance with a determination that the subset includes the first device, the first device determines a first score of a criterion associated with the voice input and receives second scores of the criterion from other devices in the subset. In accordance with a determination that the first score is higher than the second scores, the first device responds to the detected input.

Device leadership negotiation among voice interface devices

A method includes adding, by a first computing device, a first audio watermark to first speech data corresponding to playback of a first utterance including a hotword used to invoke an attention of a second computing device. The method includes outputting, by the first computing device, the playback of the first utterance corresponding to the watermarked first speech data. The second computing device is configured to receive the watermarked first speech data and determine to cease processing of the watermarked first speech data.

Hotword suppression

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for hotword detection on multiple devices are disclosed. In one aspect, a method includes the actions of receiving, by a computing device, audio data that corresponds to an utterance. The actions further include determining a likelihood that the utterance includes a hotword. The actions further include determining a loudness score for the audio data. The actions further include based on the loudness score, determining an amount of delay time. The actions further include, after the amount of delay time has elapsed, transmitting a signal that indicates that the computing device will initiate speech recognition processing on the audio data.

Hotword detection on multiple devices

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for hotword detection on multiple devices are disclosed, In one aspect, a method includes the actions of receiving audio data that corresponds to an utterance. The actions further include determining that the utterance likely includes a particular, predefined hotword. The actions further include transmitting (i) data indicating that the computing device likely received the particular, predefined hotword, (ii) data identifying the computing device, and (iii) data identifying a group of nearby computing devices that includes the computing device. The actions further include receiving an instruction to commence speech recognition processing on the audio data. The actions further include in response to receiving the instruction to commence speech recognition processing on the audio data, processing at least a portion of the audio data using an automated speech recognizer on the computing device.

Techniques disclosed herein enable training and/or utilizing speaker dependent (SD) speech models which are personalizable to any user of a client device. Various implementations include personalizing a SD speech model for a target user by processing, using the SD speech model, a speaker embedding corresponding to the target user along with an instance of audio data. The SD speech model can be personalized for an additional target user by processing, using the SD speech model, an additional speaker embedding, corresponding to the additional target user, along with another instance of audio data. Additional or alternative implementations include training the SD speech model based on a speaker independent speech model using teacher student learning.

Speaker awareness using speaker dependent speech model(s)

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for hotword trigger suppression are disclosed. In one aspect, a method includes the actions of receiving, by a microphone of a computing device, audio corresponding to playback of an item of media content, the audio including an utterance of a predefined hotword that is associated with performing an operation on the computing device. The actions further include processing the audio. The actions further include in response to processing the audio, suppressing performance of the operation on the computing device.

Recorded media hotword trigger suppression

In some implementations, processor(s) can receive an utterance from a speaker, and determine whether the speaker is a known user of a user device or not a known user of the user device. The user device can be shared by a plurality of known users. Further, the processor(s) can determine whether the utterance corresponds to a personal request or non-personal request. Moreover, and in response to determining that the speaker is not a known user of the user device and in response to determining that the utterance corresponds to a non-personal request, the processor(s) can cause a response to the utterance to be provided for presentation to the speaker at the user device response to the utterance, or can cause an action to be performed by the user device responsive to the utterance.

Multi-user authentication on a device

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a deep neural network. One of the methods includes generating a plurality of feature vectors that each model a different portion of an audio waveform, generating a first posterior probability vector for a first feature vector using a first neural network, determining whether one of the scores in the first posterior probability vector satisfies a first threshold value, generating a second posterior probability vector for each subsequent feature vector using a second neural network, wherein the second neural network is trained to identify the same key words and key phrases and includes more inner layer nodes than the first neural network, and determining whether one of the scores in the second posterior probability vector satisfies a second threshold value.

Growing community of inventors

Mountain View, CA, United States of America

Alexander Houston Gruenstein