Abstract: A method includes obtaining a set of unlabeled training samples. For each training sample in the set of unlabeled training samples generating, the method includes using a machine learning model and the training sample, a corresponding first prediction, generating, using the machine learning model and a modified unlabeled training sample, a second prediction, the modified unlabeled training sample based on the training sample, and determining a difference between the first prediction and the second prediction. The method includes selecting, based on the differences, a subset of the set of unlabeled training samples. For each training sample in the subset of the set of unlabeled training samples, the method includes obtaining a ground truth label for the training sample, and generating a corresponding labeled training sample based on the training sample paired with the ground truth label. The method includes training the machine learning model using the corresponding labeled training samples.

Abstract: A method of resizing a block storage volume for a virtual machine includes executing the virtual machine and attaching a virtual storage device to the virtual machine. The virtual storage device exposes the block storage volume on memory hardware to the virtual machine. The block storage volume includes a first storage capacity. The method also includes mounting the block storage volume to the virtual machine and resizing the block storage volume while the virtual machine continues to execute. The block storage volume is resized without attaching an additional virtual storage device to the virtual machine or mounting an additional block storage volume to the virtual machine.

Abstract: A method for a contextual assistant to use mouse pointing or touch cues includes receiving audio data corresponding to a query spoken by a user, receiving, in a graphical user interface displayed on a screen, a user input indication indicating a spatial input applied at a first location on the screen, and processing the audio data to determine a transcription of the query. The method also includes performing query interpretation on the transcription to determine that the query is referring to an object displayed on the screen without uniquely identifying the object, and requesting information about the object. The method further includes disambiguating, using the user input indication indicating the spatial input applied at the first location on the screen, the query to uniquely identify the object that the query is referring to, obtaining the information about the object requested by the query, and providing a response to the query.

Abstract: A Wi-Fi chip is configured to operate in a power saving Wi-Fi mode in which the Wi-Fi chip repeatedly checks, at a first time interval, periodic DTIM beacons transmitted at a second time interval by a wireless access point of a Wi-Fi network, wherein the first time interval is longer than the second time interval. Upon receiving a standby message, the Wi-Fi chip, during a predefined time duration, operates in a standby mode in which the Wi-Fi chip checks at a third time interval periodic DTIM beacons transmitted by the wireless access point, wherein the third time interval is shorter than the first time interval.

Abstract: This document describes foldable display support systems for thinner computing devices and improved user experience. In aspects, a device includes a foldable display coupled to the foldable display support system. The foldable display includes one or more primary folding regions of a first folding radius and one or more secondary folding regions of a second folding radius different than the first folding radius. The foldable display support system includes a support plate coupled to at least one other support plate by an adhesive layer. The support plate includes one or more flexible regions that correspond to the one or more primary folding regions of the foldable display, and the at least one other support plate includes one or more flexible regions that correspond to the one or more secondary folding regions of the foldable display.

Abstract: Various arrangements are presented for training and using a machine learning model. A first training data set may be created that has more samples but fewer dimensions than a second dataset. A second set of training data, created from the second dataset, has at least one additional dimension of data than the first set of training data. An additional dimension of data can then be simulated for the first set of training data. The simulated additional dimension of data can be incorporated with the first set of training data. A first machine learning model can be trained based on the first set of training data that comprises the simulated additional dimension of data to obtain various weights. A second machine learning model can then be trained based on the second set of training data and the obtained plurality of weights from the first trained machine learning model.

Abstract: The subject matter of this specification can be embodied in, among other things, a method that includes receiving two or more data sets each representing speech of a corresponding individual attending an internet-based social networking video conference session, decoding the received data sets to produce corresponding text for each individual attending the internet-based social networking video conference, and detecting characteristics of the session from a coalesced transcript produced from the decoded text of the attending individuals for providing context to the internet-based social networking video conference session.

Abstract: Techniques and apparatuses for a package-on-package (PoP) assembly with improved thermal management are described. In aspects, the PoP assembly includes a first IC package comprising a first IC die and a second IC package comprising a second IC die. The PoP assembly can be configured with various thermal management components that spread or dissipate heat generated by the first IC die or the second IC die of the PoP assembly. These thermal management components may include a heat spreader encapsulated within the first IC package, dummy silicon encapsulated within the first IC package, and/or a plurality of solder interconnects between the first IC package and the second IC package. By including one or more of these thermal management components, the described PoP assembly may improve thermal management of the IC packages of the PoP assembly and enable increased IC die performance or reliability over preceding assembly designs.

Abstract: An electronic device, which comprises a display element and a heat-activated adhesive film (HAF). The heat-activated adhesive film may have a planar surface complementary to and in contact with a planar surface of the display element. Another planar surface for the HAF, opposite to the planar surface of the HAF in contact with the planar surface of the display element, is complementary to and in contact with a planar surface of a laminate layer. The HAF becomes adhesive and bonds to both the display element and the laminate by being exposed to a heat source. There may also be a pressure-sensitive adhesive (PSA) layer added to the laminate layer.

Abstract: Systems and methods for audio listening devices, comprise a speaker coupled to a first housing, a sound port having a first end and a second end, wherein the first end is coupled to the first housing, and the second end is configured to be inserted in an ear canal of a person such that sound waves emitted from the speaker propagates via a secondary path to the ear canal through the sound port, active noise cancellation (ANC) components configured to generate anti-noise signals through the micro-speakers to cancel external noise, and a first microphone disposed within the sound port at the second end of the sound port such that the first microphone is configured to detect the anti-noise signal that propagates through the sound port via the secondary path and the external noise that propagates via a primary path.

Abstract: This document describes systems and techniques for a hardware-based save-and-restore controller in an SoC. The described systems and techniques can automatically save and restore access control configurations (e.g., register states) of IP subsystems during a power-down and a power-up sequence, respectively. The save operation is initiated by a local save-and-restore (L SAR) controller and performed by the IP subsystems writing the configuration values to a central save-and-restore (C-SAR) controller before powering down a power domain. The C-SAR controller saves the configuration information in a memory located in an always-on power domain. The described systems and techniques initiate, via the L SAR controller, a restore operation as part of the power-up sequence. In this way, the described systems and techniques provide scalable save-and-restore services, support a large number of power domains, and allow a variable number of access control configurations to be saved and restored.

Abstract: An acoustically-sealed electronic device assembly comprising a first housing component, a second component, and a sealing element. The first housing component may have a locking element. The second component may have a threaded channel extending at least partially along a perimeter of the second component. The threaded channel may be sized to receive the locking element. The sealing element may be positioned at an interface between the first housing component and the second component. The locking element of the first housing component may be moved from a first non-engaged position to a second engaged position within the threaded channel of the second component. The sealing element may be compressed by the first housing component and the second component to form an acoustic seal at the interface.

Abstract: This document describes techniques and systems that enable an input-mode notification for a multi-input mode. A radar field enables an electronic device to accurately determine the presence or absence of a user near the electronic device. Further, the electronic device can determine whether an application that can receive input through radar-based gestures is operating on the electronic device. Using these techniques, the electronic device can present an input-mode notification on a display of the electronic device when the user is near to alert the user that radar gesture input is available for interacting with the electronic device. This allows the device to provide the user with feedback that can educate the user about what the electronic device is capable of, and allows the user to take advantage of the additional functionality and features provided by the availability of the radar gestures.

Abstract: A method for providing autofill suggestions in a development environment includes obtaining, from a user interface executing on a user device, a user input representing source code generated within a development environment. The source code created using a particular programming language and a programming code base. The method further includes determining, using a machine learning model, at least one autofill suggestion based on the user input, the autofill suggestion continuing the source code represented by the user input. The method further includes determining, using a rule-based semantic checker configured for the particular programming language, whether the autofill suggestion is semantically correct based on the development environment and the programming code base. The method also includes, when the autofill suggestion is semantically correct, transmitting the autofill suggestion for display on the user interface of the user device.

Abstract: Techniques and apparatuses are described for adaptive selection of a network access mode by a user equipment. In aspects, a user equipment (UE) indicates, to a RAN, support for at least a first network access mode and a second network access mode and receives directions to operate in the first network access mode. While communicating in the RAN using the first network access mode, the UE detects a trigger event and determines to use the second network access mode based on at least one operational performance metric. In aspects, the UE indicates (740), to the RAN, that the UE supports the second network access mode without indicating that the UE supports the first network access mode, and transitions from the first network access mode to the second network access mode. The UE then communicates in the RAN using the second network access mode.

Abstract: A system including a power bus configured to supply power to a plurality of server racks arranged within a space of a building, a first power source connection positioned at a first side of the building and configured to supply power from a first power source to the power bus, a second power source positioned at a second side of the building different from the first side and configured to supply power from a second power source to the power bus, and a plurality of diverter switches arranged within the power bus. Each diverter switch may be configured to receive a respective control signal and, responsive to the respective control signal, redirect power within the power bus.

Abstract: Controlling voltage supplied to a load includes predicting a load current transient, generating a turbo signal in response to predicting the load current transient, and increasing, in response to the turbo signal, responsiveness of a voltage regulator supplying voltage to the load.

Abstract: A method includes receiving a system independence query requesting determination of a level of independence between a first system and a second system. The method includes obtaining a first set of time-series data including a first series of data points listed in time order and obtaining a second set of time-series data including a second series of data points listed in time order. Each data point of the first and second series of data points represents a respective system value of a feature associated with the first and second system. The method includes determining an amount of correlation between the first set of time-series data and the second set of time-series data. When the amount of correlation between the first set of time-series data and the second set of time-series data satisfies a correlation threshold, the method includes reporting that the first system and the second system are independent.

Abstract: This application is directed to a home monitoring and control system including a doorbell installed at a door of a home. The doorbell has a button configured to, upon being pressed, wirelessly initiate a first communication to indicate presence of a person at the door. The doorbell also has a camera configured to capture video data within a field of view, and a processor configured to cause a communication component to enable the first communication and wirelessly stream via a remote server the video data captured by the camera to a monitoring device associated with an occupant of the home.

Abstract: A method for securing messages includes obtaining, at a message server, a message for a user of a message service hosted by the message server. The message includes a header and the header includes a digital signature signed by an author of the message and a list of one or more recipients of the message. The method includes determining whether the digital signature by the author is valid and determining, using the list of one or more recipients, whether the user is a declared recipient of the message. When the digital signature by the author is valid and the user is the declared recipient of the message, the method includes delivering the message to a user device of the user. When the digital signature by the author is valid and the user is not the declared recipient of the message, the method includes alerting the user.

Abstract: This document describes techniques and systems that enable an interface for communicating a threshold in a camera. An electronic device recognizes an in-camera, drag gesture that triggers a camera application to switch modes from a real-time display mode (displaying real-time preview images in a viewfinder) to a buffer-display mode, which displays frames recorded in the camera buffer. During the motion of the drag gesture, the electronic device provides dynamic visual feedback indicating a relation between a drag distance of the drag gesture and a target threshold for the drag gesture. For simplicity and conciseness, the visual feedback can be combined with the virtual shutter control. After meeting the threshold, the user releases the touch input of the drag gesture and the system triggers the camera application to switch modes. This allows capture of a “missed” moment that was recorded in the camera buffer but not stored in non-volatile memory.

Abstract: A method includes receiving training data that includes unspoken textual utterances, un-transcribed non-synthetic speech utterances, and transcribed non-synthetic speech utterances. Each unspoken textual utterance is not paired with any corresponding spoken utterance of non-synthetic speech. Each un-transcribed non-synthetic speech utterance not paired with a corresponding transcription. Each transcribed non-synthetic speech utterance paired with a corresponding transcription. The method also includes generating a corresponding alignment output for each unspoken textual utterance of the received training data using an alignment model. The method also includes pre-training an audio encoder on the alignment outputs generated for corresponding to the unspoken textual utterances, the un-transcribed non-synthetic speech utterances, and the transcribed non-synthetic speech utterances to teach the audio encoder to jointly learn shared speech and text representations.

Abstract: The description relates to a device (CTRL-DEV) for controlling a dual-mode Bluetooth low energy multimedia device (DM-BLE), the dual-mode BLE multimedia device comprising a first sound system (SS1) and a second sound system (SS2) which are arranged to simultaneously stream an input multimedia stream, the first and second sound systems being respectively associated with at least one first Bluetooth multimedia device (SPK1, SPK2, SPKN) and at least one Bluetooth multimedia device (BLE-SPK1, BLE-SPK2, BLE-SPKN). The description also refers to a dual-mode Bluetooth low energy multimedia device (DM-BLE?), a method, a computer program and a non-transitory computer-readable storage medium.

Abstract: A method includes determining a cluster reliability of a computing cluster including a maximum computing capacity and representative of a reliability of the computing cluster when utilizing an entirety of the maximum computing capacity. The operations include receiving a provisioning request of the computing cluster including a threshold reliability of the computing cluster. In response to the provisioning request, determining, using the cluster reliability, a reserved computing capacity of the computing cluster based on the threshold reliability. The reserved computing capacity is less than the maximum computing capacity. Based on the reserved computing capacity and the maximum computing capacity, the operations include determining an unreserved computing capacity of the computing cluster. The operations include provisioning the computing cluster for execution of a user workload. The user workload executes on the unreserved computing capacity.

Abstract: This document describes an access point device and associated systems and methods. The techniques and systems include an access point device that includes a housing with an antenna carrier, a circuit board assembly, a heat sink, and a heat shield positioned within the housing. The housing includes a top housing member connected to a bottom housing member. The top housing member includes a concave-down top-end portion connected to a generally cylindrical vertical wall via rounded corners. The antenna carrier supports multiple antennas positioned proximate to an inner surface of the vertical wall. The heat sink is positioned between the antenna carrier and the circuit board assembly. The circuit board assembly is positioned between the heat shield and the heat sink, and the heat shield is positioned between the circuit board assembly and the bottom housing member.

Abstract: A method of detecting and responding to a visitor to a smart home environment via an electronic greeting system of the smart home environment, including determining that a visitor is approaching an entryway of the smart home environment; initiating a facial recognition operation while the visitor is approaching the entryway; initiating an observation window in response to the determination that a visitor is approaching the entryway; obtaining context information from one or more sensors of the smart home environment during the observation window; and at the end of the time window, initiating a response to the detected approach of the visitor based on the context information and/or an outcome of the facial recognition operation.

Abstract: A method for separating target speech from background noise contained in an input audio signal includes receiving the input audio signal captured by a user device, wherein the input audio signal corresponds to target speech of multiple words spoken by a target user and containing background noise in the presence of the user device while the target user spoke the multiple words in the target speech. The method also includes receiving a sequence of time markers input by the target user in cadence with the target user speaking the multiple words in the target speech, and correlating the sequence of time markers with the input audio signal to generate enhanced audio features that separate the target speech from the background noise in the input audio signal. The method also includes processing, using a speech recognition model, the enhanced audio features to generate a transcription of the target speech.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for restarting a query using a token. One of the methods includes receiving, by a computer from a requesting device, a query; determining, using a data storage system, a current result responsive to the query; generating, using the current result, a restart token that represents operations performed to determine a plurality of results responsive to the query including the current result responsive to the query and that can be used to determine a new result responsive to the query that was not included in the plurality of results responsive to the query; and providing, to the requesting device, a message that includes a) first data for the restart token that represents operations performed to determine the plurality of results responsive to the query and b) second data for the current result responsive to the query.

Abstract: Methods, systems, apparatus, including computer programs encoded on computer storage media, for reclaiming storage space in a storage environment. In one aspect, the method includes actions of aggregating data that is indicative of access to one or more data objects, determining a future storage cost associated with each of a plurality of data objects, determining an access window for each of the plurality of data objects, identifying a data object based on (i) the future storage cost that satisfies a predetermined threshold and (ii) a data object access window, providing a notification to a user device that requests feedback from a user indicating whether the data object can be deleted, and in response to receiving data that indicates that the data object can be deleted, generating an instruction to cause deletion of the data object upon the expiration of the access window.

Abstract: A method for mitigating network abuse includes obtaining a first set of network traffic messages of network traffic currently received by a network service and determining, via a first model, whether network abuse is occurring based on the first set of network traffic messages. When the network abuse is occurring, the method includes obtaining a second set of current network traffic messages. The method also includes, for each network traffic message in the second set of network traffic messages, labeling, via a second model, the network traffic message as an abusing network traffic message or a non-abusing network traffic message. The method also includes generating, via a third model, at least one network traffic rule. Each network traffic rule, when implemented, reduces an effect of the abusing network traffic messages.

Abstract: A battery pack includes a battery, a first temperature sensor configured to provide a first temperature value associated with a temperature of the battery, a heat source disposed proximate to the battery and configured to heat the battery, a second temperature sensor configured to provide a second temperature value associated with a temperature of the heat source, and a control board coupled to the first temperature sensor and the second temperature sensor, wherein the control board is configured to receive the first temperature value and the second temperature value. The control board is configured to compare the first temperature value and the second temperature value to determine a temperature gradient between the battery and the heat source and transmit an alert if the temperature gradient exceeds a first temperature gradient threshold.

Abstract: Techniques of controlling electronic devices using gestures use a wearable device on a user which translates, via a model, user movements into signals that both identify an electronic device to be controlled and a specific action to take with regard to that electronic device. The wearable device includes an inertial measurement unit (IMU) sensor and a photoplethysmography (PPG) sensor and measure six degrees of freedom (6DOF). The model is a convolutional neural network (CNN) that takes x, y, and z-acceleration signals generated by the IMU and PPG and places each acceleration component generated from each sensor in a separate channel. The CNN takes the input from each channel and generates a respective, separate model for each channel. The output at each of the stacked layers are combined in a fully connected layer to produce CNN output identifying an electronic device and a control for the electronic device.

Abstract: The technology generally relates to spatial audio communication between devices. For example, a first device and a second device may be connected via a communication link. The first device may capture audio signals in an environment through two or more microphones. The first device may encode the captured audio with direction information. The first device may transmit the encoded audio via the communication link to the second device. The second device may decode the encoded audio to be output by one or more speakers of the second device. The second device may output the decoded audio to recreate positions of the captured audio signals.

Abstract: A method includes receiving a sequence of acoustic frames as input to an automatic speech recognition (ASR) model. The method also includes generating, by a first encoder, a first higher order feature representation for a corresponding acoustic frame. The method also includes generating, by a second encoder, a second higher order feature representation for a corresponding first higher order feature representation. The method also includes generating, by a language identification (ID) predictor, a language prediction representation based on a concatenation of the first higher order feature representation and the second higher order feature representation. The method also includes generating, by a first decoder, a first probability distribution over possible speech recognition hypotheses based on a concatenation of the second higher order feature representation and the language prediction representation.

Abstract: Data are maintained in a distributed computing system that describe a graph. The graph represents relationships among items. The graph has a plurality of vertices that represent the items and a plurality of edges connecting the plurality of vertices. At least one vertex of the plurality of vertices includes a set of label values indicating the at least one vertex's strength of association with a label from a set of labels. The set of labels describe possible characteristics of an item represented by the at least one vertex. At least one edge of the plurality of edges includes a set of label weights for influencing label values that traverse the at least one edge. A label propagation algorithm is executed for a plurality of the vertices in the graph in parallel for a series of synchronized iterations to propagate labels through the graph.

Abstract: A method for time series forecasting includes receiving a time series forecasting query from a user requesting the data processing hardware to perform a plurality of time series forecasts. Each time series forecast is a forecast of future data based on respective current data. Simultaneously, for each time series forecast of the plurality of time series forecasts requested by the time series forecasting query, the method includes training a plurality of models for the respective time series forecast. The method also includes determining which model of the plurality of models best fits the respective time series forecast and forecasting the future data based on the determined best fitting model and the respective current data. The method also includes returning, to the user, the forecasted future data for each of the plurality of time series forecasts request by the timer series forecasting query.

Abstract: A method for optimal time-to-event (TTE) modeling includes obtaining a forecast request requesting performance of a TTE forecast forecasting an amount of time an event will occur after a starting point in time. The method includes obtaining a cutoff value representing an amount of time after the starting point in time that the event has not occurred. The method also includes forecasting, using an uncertainty forecasting model, the amount of time the event will occur after the starting point in time and updating the forecasted amount of time based on the cutoff value. The method also includes returning the updated forecasted amount of time the event will occur after the starting point in time.

Abstract: A method of text-only and semi-supervised training for deliberation includes receiving training data including unspoken textual utterances that are each not paired with any corresponding spoken utterance of non-synthetic speech, and training a deliberation model that includes a text encoder and a deliberation decoder on the unspoken textual utterances. The method also includes receiving, at the trained deliberation model, first-pass hypotheses and non-causal acoustic embeddings. The first-pass hypotheses is generated by a recurrent neural network-transducer (RNN-T) decoder for the non-causal acoustic embeddings encoded by a non-causal encoder. The method also includes encoding, using the text encoder, the first-pass hypotheses generated by the RNN-T decoder, and generating, using the deliberation decoder attending to both the first-pass hypotheses and the non-causal acoustic embeddings, second-pass hypotheses.

Abstract: A method includes receiving audio data corresponding to an utterance spoken by the user and captured by the user device. The utterance includes a command for a digital assistant to perform an operation. The method also includes determining, using a hotphrase detector configured to detect each trigger word in a set of trigger words associated with a hotphrase, whether any of the trigger words in the set of trigger words are detected in the audio data during the corresponding fixed-duration time window. The method also includes determining identifying, in the audio corresponding to the utterance, the hotphrase when each other trigger word in the set of trigger words was also detected in the audio data. The method also includes triggering an automated speech recognizer to perform speech recognition on the audio data when the hotphrase is identified in the audio data corresponding to the utterance.

Abstract: A method includes receiving a set of training utterances each including a non-synthetic speech representation of a corresponding utterance, and for each training utterance, generating a corresponding synthetic speech representation by using a voice conversion model. The non-synthetic speech representation and the synthetic speech representation form a corresponding training utterance pair. At each of a plurality of output steps for each training utterance pair, the method also includes generating, for output by a speech recognition model, a first probability distribution over possible non-synthetic speech recognition hypotheses for the non-synthetic speech representation and a second probability distribution over possible synthetic speech recognition hypotheses for the synthetic speech representation.

Abstract: A method for generating emotionally intelligent responses to information seeking questions includes receiving audio data corresponding to a query spoken by a user and captured by an assistant-enabled device associated with the user, and processing, using a speech recognition model, the audio data to determine a transcription of the query. The method also includes performing query interpretation on the transcription of the query to identify an emotional state of the user that spoke the query, and an action to perform. The method also includes obtaining a response preamble based on the emotional state of the user and performing the identified action to obtain information responsive to the query. The method further includes generating a response including the obtained response preamble followed by the information responsive to the query.

Abstract: A computer-implemented method includes receiving a sequence of acoustic frames corresponding to an utterance and generating a reference speaker embedding for the utterance. The method also includes receiving a target speaker embedding for a target speaker and generating feature-wise linear modulation (FiLM) parameters including a scaling vector and a shifting vector based on the target speaker embedding. The method also includes generating an affine transformation output that scales and shifts the reference speaker embedding based on the FiLM parameters. The method also includes generating a classification output indicating whether the utterance was spoken by the target speaker based on the affine transformation output.

Abstract: A method for speech conversion includes obtaining a speech conversion model configured to convert input utterances of human speech directly into corresponding output utterances of synthesized speech. The method further includes receiving a speech conversion request including input audio data corresponding to an utterance spoken by a target speaker associated with atypical speech and a speaker identifier uniquely identifying the target speaker. The method includes activating, using the speaker identifier, a particular sub-model for biasing the speech conversion model to recognize a type of the atypical speech associated with the target speaker identified by the speaker identifier.

Abstract: A multichannel neural frontend speech enhancement model for speech recognition includes a speech cleaner, a stack of self-attention blocks each having a multi-headed self attention mechanism, and a masking layer. The speech cleaner receives, as input, a multichannel noisy input signal and a multichannel contextual noise signal, and generates, as output, a single channel cleaned input signal. The stack of self-attention blocks receives, as input, at an initial block of the stack of self-attention blocks, a stacked input including the single channel cleaned input signal and a single channel noisy input signal, and generates, as output, from a final block of the stack of self-attention blocks, an un-masked output. The masking layer receives, as input, the single channel noisy input signal and the un-masked output, and generates, as output, enhanced input speech features corresponding to a target utterance.

Abstract: A method includes generating, using an audio encoder, a higher-order feature representation for each acoustic frame in a sequence of acoustic frames; generating, using a decoder, based on the higher-order feature representation, a plurality of speech recognition hypotheses, each hypotheses corresponding to a candidate transcription of an utterance and having an associated first likelihood score; generating, using an external language model, for each speech recognition hypothesis, a second likelihood score; determining, using a learnable fusion module, for each speech recognition hypothesis, a set of fusion weights based on the higher-order feature representation and the speech recognition hypothesis; and generating, using the learnable fusion module, for each speech recognition hypothesis, a third likelihood score based on the first likelihood score, the second likelihood score, and the set of fusion weights, the audio encoder and decoder trained using minimum additive error rate training in the presence of t

Abstract: A method for training hotword detection includes receiving a training input audio sequence including a sequence of input frames that define a hotword that initiates a wake-up process on a device. The method also includes feeding the training input audio sequence into an encoder and a decoder of a memorized neural network. Each of the encoder and the decoder of the memorized neural network include sequentially-stacked single value decomposition filter (SVDF) layers. The method further includes generating a logit at each of the encoder and the decoder based on the training input audio sequence. For each of the encoder and the decoder, the method includes smoothing each respective logit generated from the training input audio sequence, determining a max pooling loss from a probability distribution based on each respective logit, and optimizing the encoder and the decoder based on all max pooling losses associated with the training input audio sequence.

Abstract: A method for training a model includes obtaining a plurality of training samples. Each respective training sample of the plurality of training samples includes a respective speech utterance and a respective textual utterance representing a transcription of the respective speech utterance. The method includes training, using quantization aware training with native integer operations, an automatic speech recognition (ASR) model on the plurality of training samples. The method also includes quantizing the trained ASR model to an integer target fixed-bit width. The quantized trained ASR model includes a plurality of weights. Each weight of the plurality of weights includes an integer with the target fixed-bit width. The method includes providing the quantized trained ASR model to a user device.

Abstract: A method for detecting freeze words includes receiving audio data that corresponds to an utterance spoken by a user and captured by a user device associated with the user. The method also includes processing, using a speech recognizer, the audio data to determine that the utterance includes a query for a digital assistant to perform an operation. The speech recognizer is configured to trigger endpointing of the utterance after a predetermined duration of non-speech in the audio data. Before the predetermined duration of non-speech, the method includes detecting a freeze word in the audio data. In response to detecting the freeze word in the audio data, the method also includes triggering a hard microphone closing event at the user device. The hard microphone closing event prevents the user device from capturing any audio subsequent to the freeze word.

Abstract: A method for training a generalized automatic speech recognition model for joint acoustic echo cancellation, speech enhancement, and voice separation includes receiving a plurality of training utterances paired with corresponding training contextual signals. The training contextual signals include a training contextual noise signal including noise prior to the corresponding training utterance, a training reference audio signal, and a training speaker vector including voice characteristics of a target speaker that spoke the corresponding training utterance. The operations also include training, using a contextual signal dropout strategy, a contextual frontend processing model on the training utterances to learn how to predict enhanced speech features. Here, the contextual signal dropout strategy uses a predetermined probability to drop out each of the training contextual signals during training of the contextual frontend processing model.

Abstract: Video coding in accordance with an inter-intra prediction model may include coding an inter-prediction motion vector for a current block of a current frame, obtaining spatial block-context pixels oriented relative to the current block, generating an inter-prediction block, generating a corresponding set of reference block-context pixels oriented relative to the inter-prediction block, identifying inter-intra prediction parameters that correspond with minimizing error between the spatial block-context pixels and the reference block-context pixels, generating a prediction block for the current block by, for a current pixel of the current block, obtaining an inter-prediction pixel, determining a predictor for the current pixel using a combination of the inter-prediction pixel and the inter-intra prediction parameters, and including the predictor in the prediction block.