Abstract: Methods and systems for processing multilingual DNN acoustic models are described. An example method may include receiving training data that includes a respective training data set for each of two or more or languages. A multilingual deep neural network (DNN) acoustic model may be processed based on the training data. The multilingual DNN acoustic model may include a feedforward neural network having multiple layers of one or more nodes. Each node of a given layer may connect with a respective weight to each node of a subsequent layer, and the multiple layers of one or more nodes may include one or more shared hidden layers of nodes and a language-specific output layer of nodes corresponding to each of the two or more languages. Additionally, weights associated with the multiple layers of one or more nodes of the processed multilingual DNN acoustic model may be stored in a database.

Abstract: One embodiment of the present invention sets forth a technique for reducing flicker in image frames captured with a rolling shutter. A flicker detection and correction engine selects a first channel from a first image frame for processing. The flicker detection and correction engine subtracts each pixel value in the first channel from a corresponding pixel value in a prior image frame to generate a difference image frame. The flicker detection and correction engine identifies a first alternating current (AC) component based on a discrete cosine transform (DCT) associated with the difference image frame. The flicker detection and correction engine reduces flicker that is present in the first image frame based on the first AC component. One advantage of the disclosed techniques is that the flicker resulting from fluctuating light sources is correctly detected and reduced or eliminated irrespective of the frequency of the fluctuating light source.

Abstract: Embodiments of the present invention initially calculate a confidence score for the image environment surrounding the subject matter in order to determine the initial number of lens positions. Once the initial lens positions are determined, a sharpness score is calculated for each determined initial lens position. Using these sharpness scores, embodiments of the present invention generate a projection used to locate an estimated optimum focus position as well as to determine an estimated sharpness score at this lens position. Embodiments of the present invention then position the lens of the camera to calculate the actual sharpness score at the estimated optimum focus position, which is then compared to the estimated optimum sharpness score previously calculated. Based on this comparison, embodiments of the present invention dynamically determine whether it has a sufficient number of lens positions to determine the optimum focus position or if additional sample lens positions are needed.

Abstract: Embodiments pertain to automatic speech recognition in mobile devices to establish the presence of a keyword. An audio waveform is received at a mobile device. Front-end feature extraction is performed on the audio waveform, followed by acoustic modeling, high level feature extraction, and output classification to detect the keyword. Acoustic modeling may use a neural network or a vector quantization dictionary and high level feature extraction may use pooling.

Abstract: An image processing method including: determining whether an operation mode of a system is a control mode for still image display or a view mode for video display; and performing processing in accordance with the operation mode. Processing in the control mode includes: specifying a specific video frame image including a suitable image of a subject of image processing, from among video frame images of a video constituted of a plurality of video frame images each including an image of the subject; displaying the specific video frame image as a still image; and receiving control related to image processing from a user. Processing in the view mode includes: performing image processing with respect to each video frame image that is input, one video frame image at a time; and displaying the result of image processing for the user to view.

Abstract: Various embodiments of the present disclosure can include systems, methods, and non-transitory computer readable media configured to receive at least one database query to be executed. At least one computation graph corresponding to the at least one database query is generated. The computation graph is transformed to an optimized computation graph. The respective portions of the optimized computation graph are distributed to a plurality of distributed computing systems for execution. A result for the at least one database query is provided.

Abstract: Various embodiments of the present disclosure can include systems, methods, and non-transitory computer readable media configured to receive at least one database query to be executed. Code corresponding to the at least one database query can be generated. One or more optimizations to the generated code can be performed to produce specialized modular code. The one or more optimizations can include Just-In-Time (JIT) compilation techniques. Respective portions of the code can be distributed to a plurality of distributed computing systems for execution, wherein each of the distributed computing systems is connected to a portion of the plurality of distributed computing systems. A result for the at least one database query can be provided.

Abstract: Various embodiments can include systems, methods, and non-transitory computer readable media configured to receive at least one operation to be performed using (i) first data that is managed by a first computing system and (ii) second data that is managed by a second computing system, the operation being received through an interface provided by the computing system, and wherein the operation is based at least in part on a Structured Query Language (SQL). At least one optimization can be performed based at least in part on the operation. The operation can be executed using at least the first data and the second data. A result generated can be provided upon executing the operation through the interface provided by the computing system. The computing system, the first computing system, and the second computing system are each able to concurrently process, access, and create at least a portion of the generated result.

Abstract: Embodiments pertain to automatic speech recognition in mobile devices to establish the presence of a keyword. An audio waveform is received at a mobile device. Front-end feature extraction is performed on the audio waveform, followed by acoustic modeling, high level feature extraction, and output classification to detect the keyword. Acoustic modeling may use a neural network or Gaussian mixture modeling, and high level feature extraction may be done by aligning the results of the acoustic modeling with expected event vectors that correspond to a keyword.

Abstract: The present invention relates to social customer service and support systems integrated with social media and social networks. More particularly, the invention provides a social customer care platform system to allow customer care functions, and in particular to allow customer service agents to identify, prioritize, match and triage customer support requests that may arise through a social network and may be serviced using a social network. It manages and tracks a high-volume of customer interactions and provides for monitoring of Internet social network posts relevant to a business's products or services along with the ability to capture, monitor, filter, make sense of and respond to, in near real-time, tens of thousands of social interactions.

Abstract: One embodiment of the present invention sets forth a technique for reducing flicker in image frames captured with a rolling shutter. A flicker detection and correction engine selects a first channel from a first image frame for processing. The flicker detection and correction engine subtracts each pixel value in the first channel from a corresponding pixel value in a prior image frame to generate a difference image frame. The flicker detection and correction engine identifies a first alternating current (AC) component based on a discrete cosine transform (DCT) associated with the difference image frame. The flicker detection and correction engine reduces flicker that is present in the first image frame based on the first AC component. One advantage of the disclosed techniques is that the flicker resulting from fluctuating light sources is correctly detected and reduced or eliminated irrespective of the frequency of the fluctuating light source.

Abstract: An image processing device generates a make-up image when a make-up coating material is applied to a part of a subject appearing in a source image by processing the source image. A color range of pixels forming a coating material application layer is obtained by expanding a color range of pixels composing the part of the subject by a ratio of a color representing pixels composing the subject to a color of the make-up coating material. Values of pixels of the coating material application layer are obtained by mapping values of pixels of the source image to the color range of the coating material application layer.

Abstract: In one embodiment of the invention, there is provided a method of correcting a captured image for lens shading artifacts, comprising for a given lens determining a function L(x,y) being a lens shading correction function to be applied to images captured by the lens in order to correct for lens shading artifacts; applying a sampling technique to sample the function L(x,y) at selected points; and storing the sampled function L(x,y) in memory.

Abstract: The present disclosure is directed at a cell carrier, a stack that includes multiple cell carriers, and a method for assembling the stack. The cell carrier has a rigid backing and bus bar supports that are rigidly mounted to the rigid backing. The bus bar supports have sockets positioned to receive fasteners for securing bus bars to the bus bar supports. A battery cell that has electrodes in the form of pliable tabs can be secured to the cell carrier by, for example, adhering the cell body to the rigid backing. The cell tabs are secured between the bus bars and the bus bar supports when bus bars are fastened to the bus bar supports, and the rigidly mounted supports help prevent relative motion between the cell body and tabs. This helps prevent the cell tabs from ripping or tearing when the battery cell is subjected to vibrations during use.

Abstract: The present invention relates to social customer service and support systems integrated with social media and social networks. More particularly, the invention provides a social customer care platform system to allow customer care functions, and in particular to allow customer service agents to identify, prioritize, match and triage customer support requests that may arise through a social network and may be serviced using a social network. It manages and tracks a high-volume of customer interactions and provides for monitoring of Internet social network posts relevant to a business's products or services along with the ability to capture, monitor, filter, make sense of and respond to, in near real-time, tens of thousands of social interactions.

Abstract: In one implementation, a computer-implemented method includes receiving, at a computer system, a request to predict a next word in a dialog being uttered by a speaker; accessing, by the computer system, a neural network comprising i) an input layer, ii) one or more hidden layers, and iii) an output layer; identifying the local context for the dialog of the speaker; selecting, by the computer system and using a semantic model, at least one vector that represents the semantic context for the dialog; applying input to the input layer of the neural network, the input comprising i) the local context of the dialog and ii) the values for the at least one vector; generating probability values for at least a portion of the candidate words; and providing, by the computer system and based on the probability values, information that identifies one or more of the candidate words.

Abstract: The present disclosure is directed at a cell carrier, a stack that includes multiple cell carriers, and a method for assembling the stack. The cell carrier has a rigid backing and bus bar supports that are rigidly mounted to the rigid backing The bus bar supports have sockets positioned to receive fasteners for securing bus bars to the bus bar supports. A battery cell that has electrodes in the form of pliable tabs can be secured to the cell carrier by, for example, adhering the cell body to the rigid backing The cell tabs are secured between the bus bars and the bus bar supports when bus bars are fastened to the bus bar supports, and the rigidly mounted supports help prevent relative motion between the cell body and tabs. This helps prevent the cell tabs from ripping or tearing when the battery cell is subjected to vibrations during use.

Abstract: An injector for inserting an intraocular lens into an eye includes a lumen. The lumen can include a terminal portion at a distal end and a proximal portion juxtaposed with the terminal portion. An injector plunger can be disposed within the lumen for generating a driving force on the intraocular lens. The injector can include a lens frictional force that has a first value when the lens is at a first location within the proximal portion and a second value when the lens is at a second location within the terminal portion. The first value can be smaller than the second value. In some embodiments, the lens frictional force increases abruptly from the first value to the second value as the lens approaches the distal end of the lumen.

Abstract: An eFuse circuit may include a wordline, a first eFuse, a first logic gate, a first blowFET, and a first bitline discharge device. The first eFuse may have a first end coupled to the wordline and a second end. The first eFuse may have a first resistance when unblown and a second resistance when blown. The first logic gate may be coupled to the first end of the first eFuse. The first logic gate may be capable of driving enough current to blow the first eFuse. The first blowFET may have a source coupled to a first supply voltage, a gate coupled to a program signal, and a drain coupled to the second end of the first eFuse. The first bitline discharge device may have a gate coupled to the second end of the first eFuse, a source coupled to the first supply voltage, and a drain coupled to a first bitline.

Abstract: A speech recognition process may perform the following operations: performing a preliminary recognition process on first audio to identify candidates for the first audio; generating first templates corresponding to the first audio, where each first template includes a number of elements; selecting second templates corresponding to the candidates, where the second templates represent second audio, and where each second template includes elements that correspond to the elements in the first templates; comparing the first templates to the second templates, where comparing comprises includes similarity metrics between the first templates and corresponding second templates; applying weights to the similarity metrics to produce weighted similarity metrics, where the weights are associated with corresponding second templates; and using the weighted similarity metrics to determine whether the first audio corresponds to the second audio.