Abstract: Methods and apparatuses are provided for performing sequence to sequence (Seq2Seq) speech recognition training performed by at least one processor. The method includes acquiring a training set comprising a plurality of pairs of input data and target data corresponding to the input data, encoding the input data into a sequence of hidden states, performing a connectionist temporal classification (CTC) model training based on the sequence of hidden states, performing an attention model training based on the sequence of hidden states, and decoding the sequence of hidden states to generate target labels by independently performing the CTC model training and the attention model training.

Abstract: A semiconductor device and method of manufacturing the same are provided. The semiconductor device includes a substrate and a first gate electrode disposed on the substrate and located in a first region of the semiconductor device. The semiconductor device also includes a first sidewall structure covering the first gate electrode. The semiconductor device further includes a protective layer disposed between the first gate electrode and the first sidewall structure. In addition, the semiconductor device includes a second gate electrode disposed on the substrate and located in a second region of the semiconductor device. The semiconductor device also includes a second sidewall structure covering a lateral surface of the second gate electrode.

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: A method for tracking a maximum temperature point includes acquiring a first pair of coordinates of a maximum temperature point in a current frame of image sensed by an infrared camera, determining a rotation angle of a gimbal equipped with the infrared camera according to the first pair of coordinates of the maximum temperature point in the current frame of image and a pair of coordinates of a target position of the maximum temperature point in a subsequent frame of image, and controlling the gimbal to rotate according to the rotation angle, so as to adjust the maximum temperature point in the subsequent frame of image captured by the infrared camera to be located at the target position.

Abstract: An image presentation method includes obtaining a first image and a second image having same contents; size-processing the first image according to at least one of a target resolution, an aspect ratio of the first image, or an aspect ratio of the second image to generate a size-processed first image having the target resolution; generating a presenting image at least by combining the size-processed first image and the second image; and encoding the presenting image in a code stream and transmitting the encoded image to the display device that requires the preset resolution for display. The first and second images include a visible-light image and an infrared image. The presenting image has a preset resolution no less than a sum of the target resolution and a resolution of the second image. The size-processed first image and the second image are arranged in the presenting image without partially blocking each other.

Abstract: An image control method includes receiving, by a camera, a photographing instruction transmitted by an image display device. The camera includes a first image sensor and a second image sensor. The method further includes controlling the second image sensor to perform photographing according to the photographing instruction to obtain a display code stream and transmitting the display code stream to the image display device. The photographing instruction is used to instruct the second image sensor to photograph for a partial area of a first image using a focal length to obtain a second image. The first image is obtained by the first image sensor and displayed in a main display window of the image display device. The display code stream includes a code stream corresponding to the second image sensor.

Abstract: A method for tracking includes obtaining an infrared image and a visible image from an imaging device supported by a carrier of an unmanned aerial vehicle (UAV), combining the infrared image and the visible image to obtain a combined image, identifying a target in the combined image, and controlling at least one of the UAV, the carrier, or the imaging device to track the identified target. Combing the infrared image and the visible image includes matching the infrared image and the visible image based on matching results of different matching methods.

Abstract: A method, computer program, and computer system is provided for converting a singing first singing voice associated with a first speaker to a second singing voice associated with a second speaker. A context associated with one or more phonemes corresponding to the first singing 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 sample corresponding to the first singing voice is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

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: A control device includes a touchscreen and one or more processors. The touchscreen is configured to display an image captured by an imaging device supported by a movable object or a carrier coupled to the movable object, and receive a user input indicative of selection of a position on the touchscreen to display a selected target of the image and selection of a zoom factor for zooming in or out of the selected target. The one or more processors are configured to generate control data based on information about the user input. The control data includes instructions for the imaging device, the carrier, or the movable object to automatically control an attitude of the imaging device for positioning the selected target at or near the selected position on the touchscreen and a zoom level of the imaging device according to the user selected zoom factor.

Abstract: A method for tracking includes obtaining an infrared image and a visible image from an imaging device supported by a carrier of an unmanned aerial vehicle (UAV), combining the infrared image and the visible image to obtain a combined image, identifying a target in the combined image, and controlling at least one of the UAV, the carrier, or the imaging device to track the identified target. Combing the infrared image and the visible image includes matching the infrared image and the visible image based on matching results of different matching methods.

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: 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.

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: A computer-implemented method for tracking includes obtaining an infrared image and a visible image from an imaging device supported by a carrier of an unmanned aerial vehicle (UAV), obtaining a combined image based on the infrared image and the visible image, identifying a target in the combined image, and generating control signals for tracking the identified target using the imaging device.

Abstract: A method, computer program, and computer system is provided for converting a singing voice of a first person associated with a first speaker to a singing voice of a second person using a speaking voice of the second person associated with a second speaker. A context associated with one or more phonemes corresponding to the singing voice of a first person 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, the target acoustic frames, and a sample of the speaking voice of the second person. A sample corresponding to the singing voice of a first person is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

Abstract: A method, computer program, and computer system is provided for converting a singing voice of a first person associated with a first speaker to a singing voice of a second person using a speaking voice of the second person associated with a second speaker. A context associated with one or more phonemes corresponding to the singing voice of a first person 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, the target acoustic frames, and a sample of the speaking voice of the second person. A sample corresponding to the singing voice of a first person is converted to a sample corresponding to the second singing voice using the generated mel-spectrogram features.

Abstract: An image fusion method includes acquiring a trigger signal for flat-field correction, controlling a first image acquisition device to start a flat-field correction function to perform a flat-field correction process on the first image acquisition device according to the trigger signal for the flat-field correction, and obtaining a fused image in the flat-field correction process according to an infrared image acquired by the first image acquisition device and a visible light image acquired by a second image acquisition device. The first image acquisition device does not output infrared images during the flat-field correction process.

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 for locating a faulty photovoltaic (PV) panel includes controlling an unmanned aerial vehicle (UAV) to fly and perform image capturing, obtaining image information of the PV panel captured by a camera carried by the UAV, obtaining global positioning (GPS) information of the UAV and attitude information of the camera at a shooting time when the camera captures the image information, and, in response to determining that the image information includes fault information of the PV panel, determining a position of the PV panel according to the GPS information of the UAV and the attitude information of the camera at the shooting time.