48results about How to "Reduce hardware area" patented technology

A self-calibration circuit for capacitance mismatch is provided. The circuit comprises a sample-and-hold (S/H) circuit, a comparator, and a switch control circuit. The S/H circuit comprises a compensation capacitor array, a target capacitor, and a reference capacitor. The S/H circuit provides an output voltage, wherein the output voltage is an operation result based on the capacitance of the target capacitor and the reference capacitor, and the equivalent capacitance of the compensation capacitor array. The comparator provides a comparison signal according to whether the output voltage of the S/H circuit is positive or negative. The switch control circuit controls the equivalent capacitance of the array according to the comparison signal such that the result of the target capacitance added to the equivalent capacitance of the array gradually approximates the reference capacitance with each cycle of a clock signal.

The invention discloses an image defective pixel detection method and apparatus. The method includes: obtaining an N*M pixel block of an image, arranging all pixel points in the pixel block in sequence to obtain a pixel value sequence, calculating a median of the pixel value sequence, and obtaining a pixel median; determining a specified small pixel value and a specified large pixel value from the pixel value sequence, calculating a first difference, and determining the pixel type of a current pixel point and calculating a second difference according to the relation between the current pixel point and the pixel median; and determining whether the current pixel point is a suspected defective pixel by determining whether the quotient of the second difference and the first difference is greater than a first preset error. The defective pixel is checked based on the attribute of the defective pixel of the image, compared with the conventional defective pixel detection method, the hardware area is reduced, pixel points of the image are checked one by one, more suspected defective pixels can be discovered, and the detection efficiency of the defective pixels is improved.

Present invention relates to motion vector prediction multiplexing design method for in standard decoder, 1, firmware transmitting a avs_flag sign to MVP module, 2, adopting deblocking lower embedded subblock format, 3, taking adjacent block messages module pair frame field self adaptation sign poke, to make two different standard proceeding same processing, to realize multiplexing, 4, taking reference macroblock messages design all messages of parallel macroblock always taking from the first one reference picture in reference picture macroblock movement messages cache, in updating also saving to first reference picture, 5, the design to adjacent block data stream updating system, to make multiplexing required adjacent block information content down to the least, reducing times for reading external storage. The advantages are realizing compatibility of H,264 with AVS video coding standard MVP function, not only to ensure validity but also greatly to reduce decoder hard ware area. Said invention can reduce cache by 75 %, and realize real time decode to high resolution image.

The embodiment of the invention discloses a parallel FIR filtering method and a FIR filter, and aims to effectively reduce hardware cost and improve computing speed; the method comprises the following steps: pre-adding and merging input signals so as to obtain first to-be-processed signals, wherein the numbers of the input signals and first to-be-processed signals are bigger than 1; pre-multiplying the first to-be-processed signals with an objective filter coefficient so as to obtain second to-be-processed signals, wherein the objective filter coefficient is matched with the first to-be-processed signals; delaying and merging the second to-be-processed signals so as to obtain an output signal.

A circuit structure for implementing a multilayer artificial neural network, the circuit comprising: a plurality of memristors implementing a synaptic grid array, the memristors storing weights of the network; and a calculation and control module configured to calculate the value of weight adjustments within the network.

An apparatus of approximating backlight spread is used in a display to estimate a backlight spread image corresponding to an image after backlight spreading of a plurality of backlight sources arranged in a matrix form. An equalizer receives backlight pulse width modulation signals of the backlight sources for performing an equalization operation and generating corresponding equalization signals. A backlight seed image constructor receives the equalization signals to establish a backlight seed image. A first calculation unit calculates positions corresponding to the backlight seed image based on a backlight spread image. A second calculation unit calculates coordinates of the backlight seed image corresponding to the positions. A distance calculator calculates distance differences between the positions and coordinates of the backlight seed image. A bilinear transformation unit performs a bilinear transformation on pixels of the backlight seed image and the distance differences so as to generate the backlight spread image.

The invention discloses an image defective pixel detection method and apparatus. The method includes: defective pixel detection of an image with the bayer format is performed; a first gradient value, a second gradient value, and a third gradient value are calculated according to the position of a current pixel point and the corresponding relations between the current pixel point and surrounding pixel points; and whether the current pixel point is a suspected defective pixel is determined through comparison of the relations of the three gradient values. The defective pixels are searched based on the attribute of the defective pixels of the image, compared with the conventional defective pixel detection method, the hardware area is reduced, the pixel points of the image are checked one by one, more suspected defective pixels can be discovered, the suspected defective pixels are checked, and the efficiency of defective pixel detection is effectively improved.

The invention discloses a real-time block floating point frequency domain four-route pulse compressor and a pulse compression method thereof, and the problems that the existing pulse compression technology is long in time delay and poor in reusability are solved. The pulse compressor comprises an input data conversion module (1), a block floating point FFT module (2), a four-route matching multiplying module (3), a block floating point IFFT module (4) and an output data conversion module (5). A route of input serial data are converted by the first module (1) to four-route parallel block floating point data which are supplied to the second module (2) for FFT, after FFT, the data are supplied to the third module (3) for four-route matching multiplication, multiplication results are subjected to IFFT by the fourth module (4), and after IFFT, the data are converted by the fifth module (5) to be a route of serial fixed-point data or a route of serial floating-point data which are used as output results of pulse compression. The real-time block floating point frequency domain four-route pulse compressor has the advantages of being short in time delay and high in real-time ability, and can be used for real-time processing of radar signals.

The invention provides a method for eliminating dead pixels in an RAW-RGB image, which comprises the steps of converting R-pixels and B-pixels in an RAW-RGB image into G-pixels so as to convert the RAW-RGB image into a G-domain image, performing dead pixel judgment in the G-domain image, and then carrying out dead pixel elimination and restoration in the G-domain image. In the above process, the method can be implemented by only utilizing data of the current line where a dead pixel is located and a previous line, so that the volume of data required in the current technical scheme is greatly reduced, the time of data processing is reduced, and the possibility of real-time processing is increased. Meanwhile, the required hardware area is reduced, and the consumption of resources is lowered.

The invention discloses a Piccolo encryption algorithm hardware achieving method. The Piccolo encryption algorithm hardware can be achieved once by utilizing identical round operation in a repeated calling mode. The original r-1 round repeated calling is directly converted into r wheel repeated calling. Meanwhile, an RP-1 round displacement function is added on the r round circulation operation, and operation is conducted once to enable an output cipher to be correct. By means of the method, an original algorithm is not required to achieve again in the last rth round, the former repeated round function module can be directly multiplexed, and users only need to add the relatively simple RP-1 in the end. By means of the method, modules in the Piccolo encryption algorithm to be highly multiplexed, hardware achieving area is effectively saved, and meanwhile the encryption speed is increased.

An electronic apparatus for executing artificial intelligence algorithm is provided. The electronic apparatus includes a memory which stores input data and a plurality of second kernel data obtained from first kernel data, and a processor which obtains upscaled data in which at least a portion of the input data is upscaled by the first kernel data. The data is upscaled by performing a convolution operation on each of the plurality of second kernel data with the input data. Each of the plurality of second kernel data includes a different first kernel element from among a plurality of first kernel elements in the first kernel data.

The invention belongs to the technical field of communication and provides a method and device for determining a code phase of a ranging code, and a user terminal. The method comprises the steps of SA, obtaining discrete data; SB, obtaining carrier stripped data; SC, carrying out code matching operation according to the carrier stripped data and an initial position of sampling points, thereby obtaining a code matching result; SD, carrying out code phase search according to the code matching result and the locally copied ranging code; and SE, carrying out peak detection, if peaks are captured successfully, storing the captured peaks, adding the initial position of the sampling points and the step size x of the initial position, returning to the SC until the initial position of the sampling points is greater than the number p of the sampling points in a chip, and if the initial position of the sampling points is greater than the p, comparing p/x stored peaks, and finding the maximum peak, wherein the code phase of the ranging code is the code phase corresponding to the maximum peak. Through application of the method, the device and the user terminal, the code phase precision of the ranging code can be increased, and an area of a correlator is not increased.

The invention discloses a method for realizing moving target detection acceleration. The method comprises the steps of writing an MTD moving target detection algorithm by using C/C + + in vivalohls software; packaging the top layer into a function module; writing a test file by using C/C + + in vivalohls software; comparing the output data of the test file with the simulation output data of matlabsoftware; discriminating degree of matching, verifying correctness of a function module, aDDITION OF CONSTRAINTS AND OPTIMIZATION STRATEGY, hardware conversion is carried out to generate a hard acceleration IP core which can be synthesized by the FPGA; and automatically generating an excitation file according to the test file by using vivadohls software, automatically calling a waveform simulation tool to automatically start simulation, verifying the correctness of the hard acceleration IP core through a simulation waveform, and finally exporting a hard acceleration IP core embedded system project. According to the method, the development efficiency is improved, the project period is shortened, the hardware area is saved, and the calculation precision and the communication rate are improved.

The invention discloses an HEVC (High Efficiency Video Coding) SAO (Sample Adaptive Offset) calculation method and an HEVC SAO calculation device. The device comprises an acquisition module, a serial number calculation module, an offset value determination module, a judgment module, an addition module, and an output module. First, offset values are calculated according to original pixels acquired by the acquisition module and reference pixels corresponding to the original pixel values. Then, the judgment module judges the pixel values or offset pixel values of the original pixels output and adopted finally according to the values of code flow control command parameters acquired. Thus, SAO decoding is completed by means of two-level pipelining. The speed of SAO decoding computation is increased. The steps of the SAO decoding computation process are greatly simplified. The hardware area of an SAO computation device is reduced in hardware circuit design, and therefore, the hardware cost is lowered. Therefore, the method and the device have a broad market prospect in the field of computer chips.

The invention discloses a hardware implementation device and method for a Fruit-80 ultra-lightweight encryption algorithm. The hardware implementation device comprises a key rotation function, a nonlinear feedback shift register, a linear feedback shift register, a key stream output function and a state control unit. The key rotation function is used for providing two key correlation bits; each of the nonlinear feedback shift register and the linear feedback shift register comprises a nonlinear feedback function and a linear feedback function; the key stream output function extracts the internal states of the nonlinear feedback shift register and the linear feedback shift register in each round, and generates a key stream for encryption; and the state control unit is used for coordinating the state updating of the nonlinear feedback shift register and the linear feedback shift register in the password stage of the device. According to the hardware implementation device and method disclosed by the invention, the hardware resource occupation of the Fruit-80 ultra-lightweight sequence cipher can be reduced, and the throughput rate of the Fruit-80 ultra-lightweight sequence cipher can be improved to the greatest extent.

The invention belongs to the technical field of graphics processing chip design, and discloses a unified structure rendering shader based on a double-emission VLIW. The shader is provided with a parameter descriptor circuit for analyzing and processing input command flow information; a shading state control circuit which determines the processing contents of different pipeline stages of the unified shader according to the rendering information of the polygon to be rendered; a dual-transmit VLIW processor circuit which performs corresponding graphics task processing or general-purpose computingprocessing according to the current state and configuration parameters; a texture filter circuit which performs bilinear interpolation or trilinear interpolation filtering on the texture texture texture values read from the texture memory under the configuration of the VLIW processor circuit. The invention can reduce the complexity of hardware design and the conversion time of the data between the vertex shader and the pixel shader in the traditional architecture. The unified shader has the advantages of simple structure, small hardware area and good real-time performance.

This invention discloses a multi-purposes channels equalizer compatible with VSB/QAM. This equalizer first makes it clear that whether the VSB signals or the QAM signals, whether the symbol spacing or the fine spacing, the signal channel equalization of whether the real number or the complex number and like are for the received signals for channel equalization of the received signals. Then the corresponding filtering coefficient will be generated according to these factors and the transmission of the signals is controlled according to the various conditions. Therefore an equalizing function equalizes the VSB signals and the QAM signals. Thus, the area of the hardware for realizing the equalization can be reduced by a big margin.

The invention discloses a display device. The display device comprises a display panel, a time schedule controller, a gate enabling frequency division controller and a gate driver, and the time schedule controller is used for generating a clock signal, a gate starting signal and a gate enabling signal. The gate enabling frequency division controller is coupled to the time schedule controller and is used for generating a plurality of frequency division gate enabling signals according to the clock signal, the gate starting signal and the gate enabling signal. The gate driver is coupled to the time schedule controller, the gate enabling frequency division controller and the display panel, and is used for generating a plurality of gate output signals to the display panel according to the clock signal, the gate starting signal and the plurality of frequency division gate enabling signals. According to the invention, the display device does not have a section of redundant pre-charging and discharging, so that waste of electricity can be avoided.