Abstract: A method, computer program, and computer system is provided for synthesizing speech at one or more speeds. A context associated with one or more phonemes corresponding to a speaking voice is encoded, and the one or more phonemes are aligned to one or more target acoustic frames based on the encoded context. One or more mel-spectrogram features are recursively generated from the aligned phonemes and target acoustic frames, and a voice sample corresponding to the speaking voice is synthesized using the generated mel-spectrogram features.

Abstract: A method and apparatus include receiving a text input that includes a sequence of text components. Respective temporal durations of the text components are determined using a duration model. A spectrogram frame is generated based on the duration model. An audio waveform is generated based on the spectrogram frame. Video information is generated based on the audio waveform. The audio waveform is provided as an output along with a corresponding video.

Abstract: The electronic device includes a composite antenna including a slot antenna and a wire antenna. A first strip-shaped conductor of the slot antenna includes a first ground part, a second ground part, and a feeding part. The first ground part and the second ground part are respectively two ends of the first strip-shaped conductor. The feeding part is located between the first ground part and the second ground part. A second strip-shaped conductor of the wire antenna includes a first end and a second end. The first end of the second strip-shaped conductor is electrically connected to the first ground part. The second end of the second strip-shaped conductor is an open end.

Abstract: There is included a method and apparatus comprising computer code for generating enhanced target speech from audio data, performed by a computing device, the method comprising: receiving audio data corresponding to one or more speakers; generating estimated an target speech, an estimated noise, and an estimated echo simultaneously based on the audio data using a jointly trained complex ratio mask; predicting frame-level multi-tap time-frequency (T-F) spatio-temporal-echo filter weights based on the estimated target speech, the estimated noise, and the estimated echo using a trained neural network model; and predicting enhanced target speech based on the frame-level multi-tap T-F spatio-temporal-echo filter weights.

Abstract: The present disclosure discloses an image flicker detection method that includes the steps outlined below. An image retrieving is performed to retrieve a current image. A current variation ratio between first rows of pixels of the current image and second rows of pixels in a previous image is calculated. When both the current variation ratio and a previous variation ratio are determined to be larger than a ratio threshold, a detected flicker number is incremented. When the detected flicker number is determined to be larger than a flicker number threshold, a flicker condition is determined to occur. When the detected flicker number is determined to be not larger than the flicker number threshold, the current image becomes the previous image and the current variation ratio becomes the previous variation ratio such that a next image becomes the current image to repeat the above steps.

Abstract: An antenna structure including a slot antenna and a wire antenna is disposed, and the antenna structure is excited to generate four antenna modes: a common mode slot antenna, a differential mode slot antenna, a common mode wire antenna, and a differential mode wire antenna in a symmetric feed manner and an anti-symmetric feed manner.

Abstract: An embodiment of this application provides an electronic device, including a decoupling member, a first radiator, a second radiator, a first feed unit, a second feed unit, and a rear cover. A gap is formed between the first radiator and the second radiator, the decoupling member is indirectly coupled to the first radiator and the second radiator, and the decoupling member is disposed on a surface of the rear cover. The decoupling member does not overlap a first projection, and the first projection is a projection of the first radiator on the rear cover in a first direction. The decoupling member does not overlap a second projection, and the second projection is a projection of the second radiator on the rear cover in the first direction. The first direction is a direction perpendicular to a plane on which the rear cover is located.

Abstract: A camera optical lens includes first to fourth lenses from an object side to an image side, with first and fourth lenses having negative refractive power, and a third lens having positive refractive power, and satisfies ?3.50?f1/f??2.00; 0.55?f3/f?0.75; 5.00?d3/d4?15.00; 5.00?d5/d6?35.00; ?20.00?(R3+R4)/(R3?R4)??3.00; and ?5.00?R1/R2??2.00, where f, f1, and f3 respectively denote focal lengths of the camera optical lens, the first lens, and the third lens, d3 and d5 respectively denote on-axis thicknesses of second and third lenses, d4 and d6 respectively denote a distance between second and third lenses and a distance between third and fourth lenses, R3 and R4 respectively denote curvature radii of object side and image side surfaces of the second lens, and R1 and R2 denotes curvature radii of object side and image side surfaces of the first lens, thereby having good optical performance while meeting design requirements of a wide angle and ultra-thinness.

Abstract: The present disclosure discloses a multi-path image processing apparatus. An image merging circuit is configured to receive image frames that at least one of the image frames has a largest row number, generate redundant pixel row for each of the image frames that has a row number smaller than the largest row number such that the row number of each of the image frames equals to the largest row number, generate redundant pixel columns for each of the image frames having the number thereof determined by a size of a largest operation window, and merge each two of the image frames through the redundant columns thereof to generate a merged image frame. An image processing circuit performs image processing procedure on the merged image frame to generate a processed merged image frame, wherein at least a part of the image processing procedure is operated according to the largest operation window. An image segmentation circuit segments the processed merged image frame to generate processed image frames.

Abstract: There is included a method and apparatus comprising computer code configured to cause a processor or processors to perform generating one or more aligned inventories, wherein the one or more aligned inventories are generated using one or more word sense inventories, obtaining a word in a context sentence, determining one or more semantic equivalence scores indicating semantic similarity between the word in the context sentence and each of one or more associated glosses in the one or more aligned inventories using a semantic equivalence recognizer model, and predicting a correct sense of the word in the context sentence based on the determined one or more semantic equivalence scores.

Abstract: A method of attention-based end-to-end (A-E2E) automatic speech recognition (ASR) training, includes performing cross-entropy training of a model, based on one or more input features of a speech signal, determining a posterior probability vector at a time of a first wrong token among one or more output tokens of the model of which the cross-entropy training is performed, and determining a loss of the first wrong token at the time, based on the determined posterior probability vector. The method further includes determining a total loss of a training set of the model of which the cross-entropy training is performed, based on the determined loss of the first wrong token, and updating the model of which the cross-entropy training is performed, based on the determined total loss of the training set.

Abstract: In an approach for classifying and storing multiple layers of a file system as platform-dependent and platform-independent layers, a processor generates an initial layer of a file system. The initial layer is a platform-dependent base layer. A processor assigns one or more files associated with the initial layer with a first group identification as a first same group in a file registry for a plurality of platforms. A processor generates a new layer based on the initial layer into the file system. A processor, in response to the new layer being platform-independent, marks the new layer as platform-independent in the file registry. A processor pushes the new layer into the file registry for one of the plurality of platforms. A processor distributes one or more corresponding files from the file registry per a client request to access the file system.

Abstract: This application provides a multi-band antenna. The antenna includes a feeder and a radiating element connected to the feeder, and further includes: a first notch structure, where the first notch structure is located on a side of the radiating element and is coupled to the radiating element; and a second notch structure, where the second notch structure is located on a side of the first notch structure and far from the radiating element, and an end that is of the second notch structure and that is far from the radiating element is grounded. The first notch structure may be selectively connected to the ground or to the second notch structure. The first notch structure may be connected to the second notch structure in some embodiments using a first tuning device.

Abstract: A method, computer program, and computer system is provided for an all-deep-learning based AEC system by recurrent neural networks. The model consists of two stages, echo estimation stage and echo suppression stage, respectively. Two different schemes for echo estimation are presented herein: linear echo estimation by multi-tap filtering on far-end reference signal and non-linear echo estimation by single-tap masking on microphone signal. A microphone signal waveform and a far-end reference signal waveform are received. An echo signal waveform is estimated based on the microphone signal waveform and a far-end reference signal waveform. A near-end speech signal waveform is output based on subtracting the estimated echo signal waveform from the microphone signal waveform, and echoes are suppressed within the near-end speech signal waveform.

Abstract: In an approach for classifying and storing multiple layers of a file system as platform-dependent and platform-independent layers, a processor generates an initial layer of a file system. The initial layer is a platform-dependent base layer. A processor assigns one or more files associated with the initial layer with a first group identification as a first same group in a file registry for a plurality of platforms. A processor generates a new layer based on the initial layer into the file system. A processor, in response to the new layer being platform-independent, marks the new layer as platform-independent in the file registry. A processor pushes the new layer into the file registry for one of the plurality of platforms. A processor distributes one or more corresponding files from the file registry per a client request to access the file system.

Abstract: An image processing method is performed by a processor. The image processing method includes following operations: generating an average brightness value of each of a plurality windows in image data; generating a pixel ratio value according to the image data; generating a weight value of the each of the windows according to the pixel ratio value, a first brightness weight, and a second brightness weight; generating an adjusted brightness value according to the average brightness values of the windows and the weight values of the windows; and performing an auto exposure process according to the adjusted brightness value to generate final image data. The final image data is for a display panel to display.

Abstract: A method and apparatus are provided that analyzing sequence-to-sequence data, such as sequence-to-sequence speech data or sequence-to-sequence machine translation data for example, by minimum Bayes risk (MBR) training a sequence-to-sequence model and within introduction of applications of softmax smoothing to an N-best generation of the MBR training of the sequence-to-sequence model.

Abstract: The invention provides compounds, methods, and pharmaceutical compositions for reducing the amount or activity of Apolipoprotein E (ApoE) mRNA, and in certain embodiments reducing the amount of ApoE protein in a cell or animal, wherein reducing the amount or activity of ApoE would be beneficial.

Abstract: The invention provides compounds, methods, and pharmaceutical compositions for reducing the amount or activity of N-acylsphingosine amidohydrolase (acid ceramidase) 1 (ASAH1) mRNA, and in certain embodiments reducing the amount of ASAH1 protein in a cell or animal, wherein reducing the amount or activity of ASAH1 would be beneficial.

Abstract: Method and apparatus for automatically predicting lexical sememes using a lexical dictionary, comprising inputting a word, retrieving the word's semantic definition and sememes corresponding to the word from an online dictionary, setting each of the retrieved sememes as a candidate sememe, inputting the word's semantic definition and candidate sememe, and estimating the probability that the candidate sememe can be inferred from the word's semantic definition.