Abstract: This disclosure provides methods and apparatuses for training an acoustic model which is for implementing cross-speaker style transfer and comprises at least a style encoder. Training data may be obtained, which comprises a text, a speaker ID, a style ID and acoustic features corresponding to a reference audio. A reference embedding vector may be generated, through the style encoder, based on the acoustic features. Adversarial training may be performed to the reference embedding vector with at least the style ID and the speaker ID, to remove speaker information and retain style information. A style embedding vector may be generated, through the style encoder, based at least on the reference embedding vector being performed the adversarial training. Predicted acoustic features may be generated based at least on a state sequence corresponding to the text, a speaker embedding vector corresponding to the speaker ID, and the style embedding vector.

Abstract: A method for detecting a flow velocity of a high-temperature molten fluid can include: collecting a video stream of a high-temperature high-velocity molten fluid, decomposing the video stream into a frame image sequence sorted by time, and extracting a molten fluid Region Of Interest (ROI) from the frame image sequence, extracting a molten fluid outline of the molten fluid ROI, and extracting a characteristic block of the molten fluid outline, and obtaining the flow velocity of the molten fluid based on the characteristic block. A flow velocity detection accuracy can be improved for a molten fluid with a high temperature, a high velocity and a high glossiness.

Abstract: Systems are configured for generating spectrogram data characterized by a voice timbre of a target speaker and a prosody style of source speaker by converting a waveform of source speaker data to phonetic posterior gram (PPG) data, extracting additional prosody features from the source speaker data, and generating a spectrogram based on the PPG data and the extracted prosody features. The systems are configured to utilize/train a machine learning model for generating spectrogram data and for training a neural text-to-speech model with the generated spectrogram data.

Abstract: A data processing method, apparatus and system, for improving the security and accuracy in the data processing process. The method comprises: receiving target biometric information and verification password information of a target object sent by an acquisition institution server; recognizing, from registered biometric information of a plurality of objects, the registered biometric information matching the target biometric information; determining target account information of the target object according to the recognized registered biometric information; and sending a data processing request to a card issuing institution server corresponding to the target account information, the data processing request comprising the target account information and the verification password information, so that the card issuing institution server performs verification using the verification password information and performs data processing according to the target account information after the verification is passed.

Abstract: Systems, methods, and devices are provided for generating and using text-to-speech (TTS) data for improved speech recognition models. A main model is trained with keyword independent baseline training data. In some instances, acoustic and language model sub-components of the main model are modified with new TTS training data. In some instances, the new TTS training is obtained from a multi-speaker neural TTS system for a keyword that is underrepresented in the baseline training data. In some instances, the new TTS training data is used for pronunciation learning and normalization of keyword dependent confidence scores in keyword spotting (KWS) applications. In some instances, the new TTS training data is used for rapid speaker adaptation in speech recognition models.

Abstract: Provided is an input/output system applied to a network security defense system. A structural encoding unit and an error correction decoding unit are divided. The structure encoding unit is divided into input branch module and an input proxy module; and the error correction decoding unit is divided into an output routing module, an output proxy module, an adjudication branch module, an adjudication proxy module and a voting module. The input branch module is used for duplicating and distributing messages, the arbitration branch module is used for duplicating and distributing data, the voting module is used for performing voting, and the output routing module is used for selecting an output result from processing results of the output proxy module according to a voting result of the voting module.

Abstract: An output-decision-based negative feedback control method and system. The method includes: receiving, by an output decider, output responses of at least two heterogeneous functional equivalents, and dividing numbers corresponding to the at least two heterogeneous functional equivalents into at least one set according to the output responses; determining, by the output decider, credibility of each set according to the output responses, and sending the at least one set and the credibility corresponding to each set to a feedback controller via decision information; generating, by the feedback controller, a first scheduling policy and/or a second scheduling policy according to the decision information; and sending, by the feedback controller, the first scheduling policy to an input proxy, and/or sending a change instruction to the heterogeneous functional equivalent indicated by the second scheduling policy. The method can prevent in advance and process a heterogeneous functional equivalent that may be faulty.

Abstract: Systems, devices, and methods are provided for changing the power state of a sensor control device in an in vivo analyte monitoring system in various manners, such as through the use of external stimuli (light, magnetics) and RF transmissions.

Abstract: Provided are a security defense method and apparatus applied to a network security defense system. The method includes: using memoryless technology in a cyberspace information system, where the memoryless technology includes technology which is not affected by generalized disturbance; eliminating a memory of the cyberspace information system on an effect of random disturbance by using a redundancy and replacement mechanism; and eliminating a memory of the cyberspace information system on an effect of non-random disturbance by eliminating a memory of a program running in the cyberspace information system and/or data in the cyberspace information system. The present solution can block a memory of the cyberspace information system on an error caused by the generalized disturbance including the non-random disturbance and the random disturbance, thereby improving security of the cyberspace information system.

Abstract: Systems, devices, and methods are provided for changing the power state of a sensor control device in an in vivo analyte monitoring system in various manners, such as through the use of external stimuli (light, magnetics) and RF transmissions.

Abstract: Provided are a metal housing and a connector. The metal housing is configured to accommodate an insulation body and the metal housing includes a first half housing on which a first overlapping portion is disposed and a second half housing on which a second overlapping portion is disposed. The first overlapping portion is overlapped with and connected to the second overlapping portion when the metal housing accommodates the insulation body. With the above-mentioned structure, the metal housing is not easily deformed when being pressed and struck by external forces. The connector includes an electrical connection part and an insulation body which accommodates the electrical connection part. The connector further includes the above-mentioned metal housing which accommodates the insulation body. As the above-mentioned metal housing is used for accommodating the insulation body and the electrical connection part, the connector has better capabilities of anti-press and anti-strike.

Abstract: Described herein is a 2K clearcoat composition including, Component I including a). at least one resin having carboxylic acid groups with acid value of 50 to 400 mg KOH/g; b). at least one resin having hydroxyl functional groups with a hydroxyl value of 10 to 800 mg KOH/g; c). at least one catalyst; and d). at least one solvent, and Component II including e). at least one epoxy-functional alkoxysilane and/or its oligomer and/or its polymer; and f). at least one crosslinker reactive to hydroxyl groups. Also described herein are a substrate coated with the dried and cured 2K clearcoat composition and an automotive including the substrate.

Abstract: Disclosed embodiments include methods and systems including a receiver unit of a glucose monitoring system. The receiver is configured for communicating with a remote transmitter unit coupled with a glucose sensor. The glucose sensor generates data signals associated with a glucose level. The receiver unit includes a processor, a display, and a memory for storing instructions which, when executed by the processor: access a transmitter key associated with the remote transmitter unit; transmit a command to the remote transmitter unit after verifying the transmitter key; receive communication packets from the remote transmitter unit including a first data segment with data signals indicative of the glucose level and a second data segment with information corresponding to a remaining life of the remote transmitter unit; estimate a remaining life of the remote transmitter unit; process the data signals; and output the estimated remaining life and the processed data signals for display.

Abstract: Methods and devices to monitor an analyte in body fluid are provided. Embodiments include continuous or discrete acquisition of analyte related data from a transcutaneously positioned in vivo analyte sensor automatically or upon request from a user. The in vivo analyte sensor is coupled to an electronics unit holding a memory with instruction to cause processing circuitry to initiate a predetermined time period that is longer than a predetermined life of the sensor, during the predetermined time period, convert signals from the sensor related to glucose to respective corresponding glucose levels, without relying on any post-manufacture independent analyte measurements from a reference device, and at the expiration of the predetermined time period, disable, deactivate, or cease use of one or more feature.

Abstract: Systems are configured for generating text-to-speech data in a personalized voice by training a neural text-to-speech machine learning model on natural speech data collected from a particular user, validating the identity of the user from which data is collected, and authorizing requests from users to use the personalized voice in generating new speech data. The systems are further configured to train a machine learning model as a neural text-to-speech model with generated personalized speech data.

Abstract: A data processing method, apparatus and system, for improving the security and accuracy in the data processing process. The method comprises: receiving target biometric information and verification password information of a target object sent by an acquisition institution server; recognizing, from registered biometric information of a plurality of objects, the registered biometric information matching the target biometric information; determining target account information of the target object according to the recognized registered biometric information; and sending a data processing request to a card issuing institution server corresponding to the target account information, the data processing request comprising the target account information and the verification password information, so that the card issuing institution server performs verification using the verification password information and performs data processing according to the target account information after the verification is passed.

Abstract: Systems are configured for generating spectrogram data characterized by a voice timbre of a target speaker and a prosody style of source speaker by converting a waveform of source speaker data to phonetic posterior gram (PPG) data, extracting additional prosody features from the source speaker data, and generating a spectrogram based on the PPG data and the extracted prosody features. The systems are configured to utilize/train a machine learning model for generating spectrogram data and for training a neural text-to-speech model with the generated spectrogram data.

Abstract: A medical device includes an antenna external to a case, package, or encapsulant for the electronic systems of the medical device. In one embodiment, a diabetes infusion pump is enclosed within a metal case, the pump including a processor and a communication module for wireless communications. An antenna is disposed in the delivery tubing of the pump outside the case with an antenna feed interconnecting the external antenna with the internal communication module. In another aspect, a thin film antenna is formed on the outer surface of the case in which a physiological parameter sensor, such as a glucose sensor, is enclosed. Multiple antennas may be used for communications on different frequencies.

Abstract: Disclosed herein are methods and systems for conserving energy of a power source of an analyte monitoring device, including entering a power saving mode based on at least one of a temperature level of a power source, a level of power of a power source, or an amount of power needed by at least one component. Also disclosed herein are methods and systems for reducing noise during data transmissions to and from the analyte monitoring device.

Abstract: The present disclosure provides a method and apparatus for generating speech through neural text-to-speech (TTS) synthesis. A text input may be obtained. A phone feature of the text input may be generated. Context features of the text input may be generated based on a set of sentences associated with the text input. A speech waveform corresponding to the text input may be generated based on the phone feature and the context features.