Abstract: A method, computer system, and computer readable medium are provided for automatic speech recognition. Video data and audio data corresponding to one or more speakers is received. A minimum variance distortionless response function is applied to the received audio and video data. A predicted target waveform corresponding to a target speaker from among the one or more speakers is generated based on back-propagating the output of the applied minimum variance distortionless response function.

Abstract: A method, computer system, and computer readable medium are provided for activating speech recognition based on keyword spotting (KWS). Waveform data corresponding to one or more speakers is received. One or more direction features are extracted from the received waveform data. One or more keywords are determined from the received waveform data based on the one or more extracted features. Speech recognition is activated based on detecting the determined keyword.

Abstract: A speech recognition method, apparatus, a computer device and an electronic device for recognizing speech. The method includes receiving an audio signal obtained by a microphone array; performing a beamforming processing on the audio signal in a plurality of target directions to obtain a plurality of beam signals; performing a speech recognition on each of the plurality of beam signals to obtain a plurality of speech recognition results corresponding to the plurality of beam signals; and determining a speech recognition result of the audio signal based on the plurality of speech recognition results of the plurality of beam signals.

Abstract: A top electrode of a magnetoresistive random access memory (MRAM) device over a magnetic tunnel junction (MTJ) is formed using a film of titanium nitride oriented in a (111) crystal structure rather than a top electrode which uses tantalum, tantalum nitride, and/or a multilayer including tantalum and tantalum nitride.

Abstract: A method, computer program, and computer system for separating a target voice from among a plurality of speakers is provided. Video data associated with the plurality of speakers and audio data associated with each of the one or more speakers are received. Video feature data is extracted from the received video data. The target voice is identified from among the plurality of speakers based on the received audio data and the extracted video feature data.

Abstract: This application relates to a method of extracting an inter channel feature from a multi-channel multi-sound source mixed audio signal performed at a computing device.

Abstract: A semiconductor device and method of formation are provided. The semiconductor device includes a substrate, a first active area over the substrate, a second active area over the substrate, a graphene channel between the first active area and the second active area, and a first in-plane gate. In some embodiments, the graphene channel, the first in-plane gate, the first active area, and the second active area include graphene. A method of forming the first in-plane gate, the first active area, the second active area, and the graphene channel from a single layer of graphene is also provided.

Abstract: An operating method in a virtual environment through a wearable device is disclosed, wherein the wearable device has a motion sensor, the virtual environment has an operated object and a virtual device corresponding to the wearable device, the corresponding virtual device has a first operational data constraint, and the operated object has a second operational data constraint. The operating method comprises the following steps of: using the motion sensor to generate a motion sensed data; causing the corresponding virtual device to generate a derived data according to the motion sensed data, wherein the derived data indicates an interaction relationship between the virtual device and the operated object; and when the virtual device separated from the operated object under the interaction relationship, moving the operated object in accordance with the derived data.

Abstract: A method of manufacturing a semiconductor device includes: forming a substrate over the substrate, the substrate defining a logic region and a memory region; depositing a bottom electrode layer across the logic region and the memory region; depositing a magnetic tunnel junction (MTJ) layer over the bottom electrode layer; depositing a first conductive layer over the MTJ layer; depositing a sacrificial layer over the first conductive layer; etching the sacrificial layer in the memory region to expose the first conductive layer in the memory region while keeping the first conductive layer in the logic region covered; depositing a second conductive layer in the memory region and the logic region; patterning the second conductive layer to expose the MTJ layer in the memory region; and etching the patterned second conductive layer and the MTJ layer to form a top electrode and an MTJ, respectively, in the memory region.

Abstract: Embodiments of the present invention provide a speech signal processing model training method, an electronic device and a storage medium. The embodiments of the present invention determines a target training loss function based on a training loss function of each of one or more speech signal processing tasks; inputs a task input feature of each speech signal processing task into a starting multi-task neural network, and updates model parameters of a shared layer and each of one or more task layers of the starting multi-task neural network corresponding to the one or more speech signal processing tasks by minimizing the target training loss function as a training objective, until the starting multi-task neural network converges, to obtain a speech signal processing model.

Abstract: This application discloses a speech noise reduction method performed by a computing device. The method includes: obtaining a noisy speech signal, the noisy speech signal including a pure speech signal and a noise signal; estimating a posteriori signal-to-noise ratio and a priori signal-to-noise ratio of the noisy speech signal; determining a speech/noise likelihood ratio in a Bark domain based on the estimated posteriori signal-to-noise ratio and the estimated priori signal-to-noise ratio; estimating a priori speech existence probability based on the determined speech/noise likelihood ratio; determining a gain based on the estimated posteriori signal-to-noise ratio, the estimated priori signal-to-noise ratio, and the estimated priori speech existence probability, the gain being a frequency domain transfer function used for converting the noisy speech signal into an estimation of the pure speech signal; and exporting the estimation of the pure speech signal from the noisy speech signal based on the gain.

Abstract: A rotor-grounded motor including a stator, a rotor and a conductive plate. The stator is disposed with a bearing seat, and the bearing seat accommodates a bearing. The rotor includes a rotating shaft, and the rotating shaft is inserted in the bearing and electrically connected therewith. The conductive plate is disposed on the bearing seat, and one end of the conductive plate is connected to the bearing and the other end is electrically connected to a grounding terminal. The rotating shaft and the grounding terminal are electrically connected.

Abstract: The present disclosure provides a motor with a grounded rotor. The motor includes a stator and a rotor. The stator includes a bearing seat, a conductive plate and an elastic element. The bearing seat receives a first bearing and a second bearing. The conductive plate is disposed in the bearing seat and has at least one overlapping foot convexly formed thereon. The elastic element is arranged between the first bearing and the second bearing, and integrally formed with the at least one overlapping foot into one piece. The rotor includes a rotating shaft. The rotating shaft is inserted in the first bearing and the second bearing along the axial direction. The conductive plate and the rotating shaft are electrically connected to each other.

Abstract: A data processing method includes the following steps: generating a machine-learning parameter and obtaining a storage parameter code, wherein the storage parameter code corresponds to a storage space; receiving the machine-learning parameter and the storage parameter code, and storing the machine-learning parameter in the storage space according to the storage parameter code, and generating an event notification when the machine-learning parameter is modified; and generating a loading request according to the event notification, and the loading request is used to request the modified machine-learning parameter, wherein after the loading request is generated, the modified machine-learning parameter is downloaded from the storage space corresponding to the storage parameter code.

Abstract: The present disclosure describes an exemplary method that can prevent or reduce out-diffusion of Cu from interconnect layers to magnetic tunnel junction (MTJ) structures. The method includes forming an interconnect layer over a substrate that includes an interlayer dielectric stack with openings therein; disposing a metal in the openings to form corresponding conductive structures; and selectively depositing a diffusion barrier layer on the metal. In the method, selectively depositing the diffusion barrier layer includes pre-treating the surface of the metal; disposing a precursor to selectively form a partially-decomposed precursor layer on the metal; and exposing the partially-decomposed precursor layer to a plasma to form the diffusion barrier layer. The method further includes forming an MTJ structure on the interconnect layer over the diffusion barrier layer, where the bottom electrode of the MTJ structure is aligned to the diffusion barrier layer.

Abstract: A top electrode of a magnetoresistive random access memory (MRAM) device over a magnetic tunnel junction (MTJ) is formed using a film of titanium nitride oriented in a (111) crystal structure rather than a top electrode which uses tantalum, tantalum nitride, and/or a multilayer including tantalum and tantalum nitride.

Abstract: This application discloses a sound acquisition component array, including: two first sound acquisition components, two second sound acquisition components, and two third sound acquisition components. The two second sound acquisition components are located at a first side of a line connecting the two first sound acquisition components, and the two third sound acquisition components are located at a second side of the connecting line that is opposite to the first side of the connecting line; the two second sound acquisition components are symmetrical about a perpendicular bisector of the connecting line, and the two third sound acquisition components are symmetrical about the perpendicular bisector; and a distance between the two first sound acquisition components, a distance between the two second sound acquisition components, and a distance between the two third sound acquisition components are respectively different from one another along a direction defined by the connecting line.

Abstract: The invention generally relates to solid forms of berberine ursodeoxycholate, and to pharmaceutical compositions, and methods of preparation and therapeutic use thereof. In particular, the invention relates to solid forms of berberine ursodeoxycholate and pharmaceutical compositions thereof useful in treating and/or preventing various diseases or disorders, including metabolic disorders, heart diseases, neurodegenerative diseases and liver diseases.

Abstract: Methods of forming magnetic tunnel junction (MTJ) memory cells used in a magneto-resistive random access memory (MRAM) array are provided. A pre-clean process is performed to remove a metal oxide layer that may form on the top surface of the bottom electrodes of MTJ memory cells during the time the bottom electrode can be exposed to air prior to depositing MTJ layers. The pre-clean processes may include a remote plasma process wherein the metal oxide reacts with hydrogen radicals generated in the remote plasma.

Abstract: Face anti-counterfeiting detection methods and systems, electronic devices, and computer storage media include: obtaining an image or video to be detected containing a face; extracting a feature of the image or video to be detected, and detecting whether the extracted feature contains counterfeited face clue information; and determining whether the face passes the face anti-counterfeiting detection according to a detection result.