Abstract: The present disclosure provides an operating method of a speech synthesis system, which includes, inputting a first text and a first speech for the first text, and a second text and a second speech for the second text; generating a speech synthesis model trained by applying the first and second texts and the first and second speeches to curriculum learning; and outputting a target synthesis speech corresponding to a target text based on the speech synthesis model when inputting the target text for speech output, and the generating of the speech synthesis model includes generating a concatenation text in which the first and second texts are concatenated and a concatenation speech in which the first and second speeches are concatenated, and adding the concatenation text and the concatenation speech to the speech synthesis model when an error rate is smaller than a set reference rate when learning-concatenating the concatenation text and the concatenation speech.

Abstract: Disclosed is a processor implemented method that includes calculating a quantization error for each channel of a neural network using activation data output from a first layer of the neural network and a quantization scale of a second layer connected to the first layer, calculating a final loss using a regularization loss term determined based on the quantization error for each channel, and updating a batch norm parameter of the first layer in a direction to decrease the final loss.

Abstract: A material layer manufacturing method is provided. The material layer manufacturing method may comprise the steps of: preparing a substrate having a base pattern formed thereon; providing a first precursor on the substrate having the base pattern formed thereon, in a state where a first voltage is applied to the base pattern; and providing a second precursor on the substrate having the first precursor provided thereon, in a state where a second voltage is applied to the base pattern, to form, on the substrate having the base pattern formed thereon, a material layer resulting from the reaction of the first precursor with the second precursor.

Abstract: The present invention relates to a robot. A robot according to one embodiment of the present invention comprises: a body having a set volume: and a driving module provided in the left region and the right region of the body, respectively, and rotatably connected to the body by a rotation shaft, wherein the driving module comprises: a rotating member connected to the rotating shaft; a first driving part extending from one side of the rotating member slantwise by a first inclination angle with respect to the rotating shaft in a direction away from the rotating shaft; and a second driving part extending from the one side of the rotating member slantwise by a second inclination angle with respect to the rotating shaft in the direction away from the rotating shaft.

Abstract: Disclosed are a method and an apparatus for detecting a voice end point by using acoustic and language modeling information to accomplish strong voice recognition. A voice end point detection method according to an embodiment may comprise the steps of: inputting an acoustic feature vector sequence extracted from a microphone input signal into an acoustic embedding extraction unit, a phonemic embedding extraction unit, and a decoder embedding extraction unit, which are based on a recurrent neural network (RNN); combining acoustic embedding, phonemic embedding, and decoder embedding to configure a feature vector by the acoustic embedding extraction unit, the phonemic embedding extraction unit, and the decoder embedding extraction unit; and inputting the combined feature vector into a deep neural network (DNN)-based classifier to detect a voice end point.

Abstract: The present invention relates to a color-changing sensor for gas detection and a method for manufacturing same and, more particularly, to a color-changing sensor that can be used for detecting hydrogen gas, in which when the sensor comes into contact with hydrogen gas, color conversion occurs so that it is possible to easily check the presence of hydrogen gas, and when the sensor is not in contact with hydrogen gas by removal of hydrogen gas or the like, the sensor may exhibit a reversible characteristic of being restored to an original color and an irreversible characteristic of not being restored to an original color.

Abstract: A voice signal estimation apparatus using attention mechanism according to an embodiment may comprise a microphone encoder that receives a microphone input signal including an echo signal, and a user's voice signal, converts the microphone input signal into first input information, and outputs the converted first input information, a far-end signal encoder that receives a far-end signal, converts the far-end signal into second input information, and outputs the converted second input information, an attention unit outputting weight information by applying an attention mechanism to the first input information and the second input information, a pre-learned first artificial neural network with third input information, which is the sum information of the weight information and the second input information, as input information, and with first output information including mask information for estimating the voice signal from the second input information as output information and a voice signal estimator outputtin

Abstract: A selector according to an embodiment of the present disclosure includes a first electrode; a second electrode disposed opposite to the first electrode; an ion supply layer disposed between the first electrode and the second electrode to be on the side of the first electrode and doped with a metal, wherein the doped metal diffuses toward the second electrode; a switching layer disposed between the first electrode and the second electrode to be on the side of the second electrode, wherein the doped metal diffuses from the ion supply layer into the switching layer so that metal concentration distribution inside the switching layer is changed to generate metal filaments; and a diffusion control layer inserted between the ion supply layer and the switching layer, wherein the diffusion control layer serves to adjust electrical characteristics related to the generated metal filaments as the amount of the diffusing metal is adjusted in proportion to a thickness of the diffusion control layer.

Abstract: An organic light-emitting display device includes quantum dots and an RGB color filter layer having quantum dots and thus is capable of removing 100% of interference among red, green, and blue color filters.

Abstract: Disclosed are a BCTZ-based lead-free piezoelectric ceramic composition and a method of preparing a BCTZ-based lead-free piezoelectric ceramic which provide high piezoelectric properties and speaker characteristics. The method of preparing a lead-free piezoelectric ceramic includes preparing a mixture in which copper oxide is added to a BCTZ-based base material and sintering the mixture at a temperature of 1300° C. to 1600° C. for 1 to 5 hours, wherein a weight ratio of the copper oxide in the mixture is 0.6 wt %.

Abstract: Provided is a rack for a bicycle or a kick scooter that is a single device allowing either of a bicycle or a kick scooter to be parked. The rack includes a pair of support members formed to extend upward from the ground and disposed to be parallel with and spaced apart from each other, a rotating shaft coupled to the support members, a parking member interposed between the pair of support members and axially coupled to the rotating shaft so as to be rotatable, and a protruding member disposed to be spaced apart from a front of the parking member that faces the bicycle or the kick scooter and having both ends bent toward the parking member to be fixed to the parking member.

Abstract: An anode active material comprising a transition metal oxide is provided. A method for preparing the anode active material may comprise the steps of preparing a first transition metal oxide source and a second transition metal oxide source; providing the first transition metal oxide source and the second transition metal oxide source for a secondary alcohol to prepare a base source; providing a hydrolysis catalyst for the base source and inducing a sol-gel reaction to prepare a transition metal oxide precursor; and subjecting the transition metal oxide precursor to heat treatment in a nitrogen environment to prepare an anode active material containing a transition metal oxide.

Abstract: [SUMMARY] An integrated noise and echo signal removal device using parallel deep neural network according to an embodiment comprises a microphone encoder that receives a microphone input signal including an echo signal, and a speaker's voice signal, converts the microphone input signal into first input information, and outputs the converted first input information, a far-end signal encoder that receives a far-end signal, converts the far-end signal into second input information, and outputs the converted second input information, a pre-learned second artificial neural network having a third input information, which is the sum of the first input information and the second input information, as input information, and having an estimated echo signal obtained by estimating the echo signal from the second input information as output information, a pre-learned third artificial neural network having the third input information as input information and having an estimated noise signal obtained by estimating the noise

Abstract: A porous polymer structure having a smooth surface and a porous structure embedded below the surface can be simply produced by using a hydrophobic substrate with a steam spraying process, and the porous polymer structure can exhibit very excellent low-adhesive property due to the smooth surface and the porous structure embedded below the surface. It is possible to realize excellent low-adhesive property even without using a surface modifier or a lubricant. Since the structure shows flexibility, it can even be attached to a curved surface, and can be usefully applied to various industries due to a possible adjustment of the surface characteristic.

Abstract: The present invention relates to a novel use of NCKAP1 gene in neurodegenerative diseases. More specifically, the present invention relates to a marker composition for predicting the prognosis of a neurodegenerative disease, comprising a NCKAP1 protein or a gene encoding same, a composition and a kit for predicting the prognosis of a neurodegenerative disease, which comprises a formulation for measuring the level of the protein or an mRNA of the gene encoding same, and a pharmaceutical composition for preventing or treating neurodegenerative disease, comprising the protein or the gene encoding same as an active ingredient.

Abstract: Provided is a water electrolysis catalyst with a core-shell structure, which has a vanadium-doped cobalt nitride (V—Co4N) core; and a cobalt-nickel phosphate (CoNiPOx, x is a natural number) shell.

Abstract: The present disclosure relates to a two-phase Byzantine consensus method and system for state machine replication. A consensus method which is performed by a consensus system includes performing, by all of normal processes, consensus for a sequence number proposed by a process designated to be primary with respect to a client request according to an agreement protocol; and replacing the primary when the primary is suspicious of having a Byzantine fault according to a view change protocol.

Abstract: Provided is a method of compressing an occupancy map of a three-dimensional (3D) point cloud, and more specifically, a method of compressing an occupancy map of a point cloud in which an occupancy map image of a point cloud existing in a 3D space is compressed based on a compression quality or a patch-by-patch inspection method of the occupancy map image so that compression distortion is minimized when reconstructing the compressed occupancy map image so as to remarkably improve the quality of a reconstructed occupancy map image.

Abstract: An image encoding/decoding method using a pixel value range constituting an image is disclosed, wherein the image decoding method using a pixel value range constituting an image comprises the steps of: receiving a bitstream; acquiring information of a pixel value range forming a first unit image included in the received bitstream; and decoding the first unit image on the basis of the acquired information of the pixel value range. Therefore, compression efficiency can be improved in an image encoding or decoding process.

Abstract: Provided are a method for manufacturing a single-crystal semiconductor layer. The method of manufacturing the single crystalline semiconductor layer includes performing a unit cycle multiple times, wherein the unit cycle includes a metal precursor pressurized dosing operation in which a metal precursor is adsorbed on a surface of a single crystalline substrate by supplying the metal precursor onto the single crystalline substrate while an outlet of a chamber in which the single crystalline substrate is loaded is closed such that a reaction pressure in the chamber is increased; a metal precursor purge operation; a reactive gas supplying operation in which a reactive gas is supplied into the chamber to cause a reaction of the reactive gas with the metal precursor adsorbed on the single crystalline substrate after the metal precursor purge operation; and a reactive gas purge operation.