Abstract: A microphone system of the invention is applicable to an electronic device comprising an adjustable mechanism that causes a change in geometry of a microphone array. The microphone system comprises the microphone array, a sensor and a beamformer. The microphone array comprises multiple microphones that detect sound and generate multiple audio signals. The sensor detects a mechanism variation of the electronic device to generate a sensing output. The beamformer is configured to perform a set of operations comprising: performing a spatial filtering operation over the multiple audio signals using a trained model based on the sensing output, one or more first sound sources in one or more desired directions and one or more second sound sources in one or more undesired directions to generate a beamformed output signal originated from the one or more first sound sources.

Abstract: An in-memory computing memory device is disclosed. The memory device comprises an array of memory cells, a plurality of word lines, a plurality of bit lines, (M+1) input circuits, a wordline driver and an evaluation circuitry. The array is divided into (M+1) lanes and each lane comprises P memory cell columns and an input circuit. The input circuit in each lane charges a predefined bit line with a default amount of charge proportional to an input synapse value and then distributes the default amount of charge to the other second bit lines with a predefined ratio based on a constant current. The evaluation circuitry couples a selected number of the bit lines to an accumulate line and convert an average voltage at the accumulate line into a digital value in response to a set of (M+1) input synapse values and the activated word line.

Abstract: An integrated circuit applied in a deep neural network is disclosed. The integrated circuit comprises at least one processor, a first internal memory, a second internal memory, at least one MAC circuit, a compressor and a decompressor. The processor performs a cuboid convolution over decompression data for each cuboid of an input image fed to any one of multiple convolution layers. The MAC circuit performs multiplication and accumulation operations associated with the cuboid convolution to output a convoluted cuboid. The compressor compresses the convoluted cuboid into one compressed segment and store it in the second internal memory. The decompressor decompresses data from the second internal memory segment by segment to store the decompression data in the first internal memory. The input image is horizontally divided into multiple cuboids with an overlap of at least one row for each channel between any two adjacent cuboids.

Abstract: An in-memory computing memory device is disclosed. The memory device comprises an array of memory cells, a plurality of word lines, a plurality of bit lines, (M+1) input circuits, a wordline driver and an evaluation circuitry. The array is divided into (M+1) lanes and each lane comprises P memory cell columns and an input circuit. The input circuit in each lane charges a predefined bit line with a default amount of charge proportional to an input synapse value and then distributes the default amount of charge to the other second bit lines with a predefined ratio based on a constant current. The evaluation circuitry couples a selected number of the bit lines to an accumulate line and convert an average voltage at the accumulate line into a digital value in response to a set of (M+1) input synapse values and the activated word line.

Abstract: Calibration methods for calibrating image capture devices of an around view monitoring (AVM) system mounted on vehicle are provided, the calibration method including: extracting local patterns from images captured by each image capture device, wherein each local pattern is respectively disposed at a position within the image capturing range of one of the image capture devices; acquiring an overhead-view (OHV) image from OHV point above vehicle, wherein the OHV image includes first patterns relative to the local patterns for the image capture devices; generating global patterns from the OHV image using the first patterns, each global pattern corresponding to one of the local patterns; matching the local patterns with the corresponding global patterns to determine camera parameters and transformation information corresponding thereto for each image capture device; and calibrating each image capture device using determined camera parameters and transformation information corresponding thereto so as to generate A

Abstract: A dual-camera device is provided. The dual-camera device includes wide and telephoto imaging sections with respective lens/sensor combinations, and a processor. The wide and telephoto imaging sections provide wide image data and telephoto image data, respectively. At least one misalignment error exists between the wide and telephoto imaging sections. The processor generates an output image provided with a smooth transition when switching between a lower zooming factor and a higher zooming factor. The processor warps the wide image data using a portion of the misalignment error to generate base wide image data, and warps the telephoto image data using the remaining portion of the misalignment error to generate base telephoto image data. The processor generates the output image using the base wide image data at the lower zooming factor, and generates the output image using the base telephoto image data at the higher zooming factor.

Abstract: Calibration methods for calibrating image capture devices of an around view monitoring (AVM) system mounted on vehicle are provided, the calibration method including: extracting local patterns from images captured by each image capture device, wherein each local pattern is respectively disposed at a position within the image capturing range of one of the image capture devices; acquiring an overhead-view (OHV) image from OHV point above vehicle, wherein the OHV image includes first patterns relative to the local patterns for the image capture devices; generating global patterns from the OHV image using the first patterns, each global pattern corresponding to one of the local patterns; matching the local patterns with the corresponding global patterns to determine camera parameters and transformation information corresponding thereto for each image capture device; and calibrating each image capture device using determined camera parameters and transformation information corresponding thereto so as to generate A

Abstract: A method and apparatus for three-dimensional video coding using the virtual depth information are disclosed. For a current texture block in the dependent view, the method incorporating the present invention first derives an estimated disparity vector to locate a corresponding texture block in a coded view. A collocated depth block in the coded view collocated with the corresponding texture block in the coded view is identified and used to derive the virtual depth information. One aspect of the present invention addresses derivation process for the estimated disparity vector. Another aspect of the present invention addresses the usage of the derived virtual depth information.

Abstract: A dual-camera device is provided. The dual-camera device includes wide and telephoto imaging sections with respective lens/sensor combinations, and a processor. The wide and telephoto imaging sections provide wide image data and telephoto image data, respectively. At least one misalignment error exists between the wide and telephoto imaging sections. The processor generates an output image provided with a smooth transition when switching between a lower zooming factor and a higher zooming factor. The processor warps the wide image data using a portion of the misalignment error to generate base wide image data, and warps the telephoto image data using the remaining portion of the misalignment error to generate base telephoto image data. The processor generates the output image using the base wide image data at the lower zooming factor, and generates the output image using the base telephoto image data at the higher zooming factor.

Abstract: An electronic apparatus includes at least two camera devices and a processing device. The processing device determines a first distance to a surface formed by the two camera devices and a second distance to the surface in response to detecting an object positioned at a first time by the two camera devices, and determines a third distance from the object positioned at a second time to the surface, wherein the second time is later than the first time, and the third distance is longer than the first distance and shorter than the second distance. Also, the processing device determines a depth in a virtual space corresponding to the object positioned at the second time according to the first distance, the second distance, and the third distance.

Abstract: An apparatus and method for temporal motion vector prediction for a current block in a picture are disclosed. In the present method, one temporal block in a first reference picture in a first list selected from a list group comprising list 0 and list 1 is determined. When the determined temporal block has at least one motion vector, a candidate set is determined based on the motion vector of the temporal block. The temporal motion vector predictor or temporal motion vector predictor candidate or temporal motion vector or temporal motion vector candidate for the current block is determined from the candidate set by checking a presence of a motion vector pointing to a reference picture in a first specific list in said at least one motion vector, wherein the first specific list is selected from the list group based on a priority order.

Abstract: A method and apparatus for deriving MV/MVP (motion vector or motion vector predictor) or DV/DVP (disparity vector or disparity vector predictor) associated Skip mode, Merge mode or Inter mode for a block of a current picture in three-dimensional (3D) video coding are disclosed. The 3D video coding may use temporal prediction and inter-view prediction to exploit temporal and inter-view correlation. MV/DV prediction is applied to reduce bitrate associated with MV/DV coding. The MV/MVP or DV/DVP for a block is derived from spatial candidates, temporal candidates and inter-view candidates. For the inter-view candidate, the position of the inter-view co-located block can be located using a global disparity vector (GDV) or warping the current block onto the co-located picture according to the depth information. The candidate can also be derived as the vector corresponding to warping the current block onto the co-located picture according to the depth information.

Abstract: A method and apparatus for texture image compression in a 3D video coding system are disclosed. Embodiments according to the present invention derive depth information related to a depth map associated with a texture image and then process the texture image based on the depth information derived. The invention can be applied to the encoder side as well as the decoder side. The encoding order or decoding order for the depth maps and the texture images can be based on block-wise interleaving or picture-wise interleaving. One aspect of the present invent is related to partitioning of the texture image based on depth information of the depth map. Another aspect of the present invention is related to motion vector or motion vector predictor processing based on the depth information.

Abstract: A method for deriving a motion vector predictor (MVP) receives motion vectors (MVs) associated with reference blocks of the current block. The method determines at least one first spatial search MV associated with a first MV searching order and at least one second spatial search MV associated with a second MV searching order for each neighboring reference block. Then, the method determines whether a first available-first spatial search MV exists for said at least one neighboring reference block according to the first MV searching order, and provides the first available-first spatial search MV as a spatial MVP for the current block. Finally, the method determines whether a first available-second spatial search MV exists for said at least one neighboring reference block according to the second MV searching order only if none of first spatial search MVs for said at least one neighboring reference block is available.

Abstract: An exemplary method for determining a control output in a control domain includes: obtaining a control input of an input domain, wherein the control input includes a previous input value and a current input value; and dynamically adjusting a control resolution setting, and converting the control input of the input domain into the control output in the control domain according to the control resolution setting, wherein the control output includes a previous output value and a current output value corresponding to the previous input value and the current input value, respectively, the control resolution setting for the current input value is determined according to at least the previous input value, and the current output value is identical to the previous output value when the current input value and the previous input value are generated in response to a same user input.

Abstract: A method and apparatus for deriving a motion vector predictor (MVP) for a motion vector (MV) of a current block of a current picture in Inter, or Merge, or Skip mode. The method selects a co-located block corresponding to a co-located picture, receives one or more reference motion vectors (MVs) of one or more co-located reference blocks associated with the co-located block, and determines a search set that includes one or more search MVs associated with said one or more co-located reference blocks. The method further determines a search order for the search set having an order associated with a reference list attribute for a first search MV and a second search MV associated with an identical co-located reference block. Finally, the method determines the MVP for the current block based on the search set and the search order, and provides the MVP for the current block.

Abstract: An apparatus and method of deriving a motion vector predictor (MVP) for a current MV of a current block in Inter, Merge or Skip mode are disclosed based on motion vector (MV) attribute search. The system determines first MV attribute search comprising whether a given MV pointing to the target reference picture in the given reference list, or whether the given MV pointing to the target reference picture in other reference list, and determines second MV attribute search comprising whether the given MV pointing to other reference pictures in the given reference list, or whether the given MV pointing to the other reference pictures in the other reference list. The MVP for the current block is then determined from the neighboring blocks according to a search order.

Abstract: The invention provides a motion prediction method. First, a plurality of candidate units corresponding to a current unit of a current frame is determined. A plurality of motion vectors of the candidate units is then obtained. A plurality of scaling factors of the candidate units is then calculated according to a plurality of respective temporal distances depending on a plurality of reference frames of the motion vectors. The motion vectors of the candidate units are then scaled according to the scaling factors to obtain a plurality of scaled motion vectors. The scaled motion vectors are ranked, and a subset of highest ranking motion vectors are identified to be included in a candidate set. Finally, a motion vector predictor for motion prediction of the current unit is then selected from the candidate units.

Abstract: A method and apparatus for deriving a motion vector predictor (MVP) for a motion vector (MV) of a current block of a current picture in Inter, or Merge, or Skip mode. The method selects a co-located block corresponding to a co-located picture, receives one or more reference motion vectors (MVs) of one or more co-located reference blocks associated with the co-located block, and determines a search set that includes one or more search MVs associated with said one or more co-located reference blocks. The method further determines a search order for the search set having an order associated with a reference list attribute for a first search MV and a second search MV associated with an identical co-located reference block. Finally, the method determines the MVP for the current block based on the search set and the search order, and provides the MVP for the current block.

Abstract: A method and apparatus for deriving a temporal motion vector predictor (MVP) are disclosed. The MVP is derived for a current block of a current picture in Inter, or Merge, or Skip mode based on co-located reference blocks of a co-located block and a flag is used to indicate the co-located picture. More than one co-located reference blocks can be used to derive the temporal MVP and the co-located reference blocks can be selected from the co-located block as well as neighboring blocks of the co-located block. A search set comprises search motion vectors associated with the co-located reference block(s) is formed. The search motion vector (MV) corresponding to the co-located reference block in the same reference list is searched before the search MV in a different reference list. Various schemes to accommodate implicit method of deriving co-located picture are also disclosed.