128 results about "Reconstruction filter" patented technology

The present disclosure is directed to an augmented reality surgical system for viewing an augmented image of a region of interest during a surgical procedure. The system includes an image capture device that captures an image of the region of interest. A controller receives the image and applies at least one image processing filter to the image. The image processing filter includes a spatial decomposition filter that decomposes the image into spatial frequency bands. A temporal filter is applied to the spatial frequency bands to generate temporally filtered bands. An adder adds each band spatial frequency band to a corresponding temporally filtered band to generate augmented bands. A reconstruction filter generates an augmented image by collapsing the augmented bands. A display displays the augmented image to a user.

An apparatus and method of training a convolutional neural network (CNN) are provided. A method of training a CNN including a plurality of convolution layers stored in a memory involves approximating, using a processor, a convolution layer among the plurality of convolution layers using a low-rank approximation; reducing the number of output reconstruction filters of the approximated convolution layer; and modifying a structure of the CNN based on an approximation result and the reduced number of output reconstruction filters.

The invention discloses a same-time same-frequency full duplex multi-tap radio frequency self-interference offset system and method. A wireless unit receiving end outputs self-interference signals, self-interference offset is not conducted at first, after being processed by a radio frequency receiving channel and an analog-digital converter, the self-interference signals are sent to a pattern recognition module to generate an adjustment value, multiple taps are controlled to interfere in a reconstruction filter to reconstruct sending end radio frequency signals in a radio frequency mode, radio frequency self-interference offset is conducted on output reconstruction signals through a summator, signals after the radio frequency self-interference offset are sent to an adjustment algorithm module again through the radio frequency receiving channel and the analog-digital converter to obtain digital signals which are sent to the adjustment algorithm module, an adjusting value needed for adjusting the time delay zero is obtained, radio frequency interference reconstruction is further controlled, controlling over the radio frequency interference reconstruction is carried out repeatedly according to the method, and self-interference offset is achieved. According to the same-time same-frequency full duplex multi-tap radio frequency self-interference offset system and method, joint control of the pattern recognition module and the adjustment algorithm module is used, same-time same-frequency-radio self-interference can be removed effectively, the frequency spectrum utilization rate is improved, the radio frequency self-interference offset performance is improved, and the working stability of communication equipment is guaranteed.

The invention discloses a self-adoptive correcting device of mismatch error of time-interleaved analog-digital converter, comprising an M channel TIADC, a signal recombination, a digital reference signal memorizer, a simulated reference signal generator, a self-adaptive reconstruction filter bank, a clock generation circuit and a subtraction device. Signals after passages are reconstructed are used to correct each passage instead of single correction on each passage, thereby solving the problem that when an input signal bandwidth is larger than the Nyquist frequency of each passage ADC, the time error can not be corrected due to aliasing. Each self-adoptive reconstruction filter is divided into a plurality of sub-filters for concurrent working, thereby not improving the requirement of thetreatment speed for a self-adoptive correcting filter while realizing the effect of signal recombination and ensuring the practicability of the hardware of the structure of the invention. A digital reference signal is internally installed in the device and is taken as a target to carry out the self-adoptive correction, pre-measuring or calculating a passage mismatch error is not needed, and the source of the error is not needed to be discriminated so that various mismatch errors can be corrected.

Sound source separation, without permutation, using convolutional mixing independent component analysis based on a priori knowledge of the target sound source is disclosed. The target sound source can be a human speaker. The reconstruction filters used in the sound source separation take into account the a priori knowledge of the target sound source, such as an estimate the spectra of the target sound source. The filters may be generally constructed based on a speech recognition system. Matching the words of the dictionary of the speech recognition system to a reconstructed signal indicates whether proper separation has occurred. More specifically, the filters may be constructed based on a vector quantization codebook of vectors representing typical sound source patterns. Matching the vectors of the codebook to a reconstructed signal indicates whether proper separation has occurred. The vectors may be linear prediction vectors, among others.

A method for improving disease diagnosis using contrast enhancement presentation comprising: providing an input digital diagnostic image; applying a decomposition filter bank to the input digital diagnostic image; constructing a tone scale curve from the input digital diagnostic image; applying said tone scale curve to the input digital diagnostic image to produce a tone-scaled image; applying a decomposition filter bank to the tone-scaled image; generating the contrast weight control signals from the input digital diagnostic image by extracting the high contrast edge signals at the coarse scale; adjusting the decomposition outputs from both the input image and the tone-scaled image according to the contrast weight control signals; and applying a reconstruction filter bank to the adjusted signals to produce a contrast enhancement presentation output image.

An image is reconstructed based upon a filtered backprojection (FBP) technique using a reconstruction filter which is customized or shaped by parameters including a weight value that is determined for weighing projection data according to a predetermined noise model. The weight value is determined based upon rays or views in the projection data. A regularization term is also combined with the noise-weighted FBP.

An integrated test core for mixed-signal circuits comprises a periodic waveform generator capable of generating arbitrary band-limited waveforms for excitation purposes and a waveform digitizer for extracting an arbitrary waveform from the test circuit's analog response signal. The digitized response may be tested and measured using DSP techniques. Preferably, the waveform generator and digitizer are synchronously controlled. The core is a nearly all digital implementation with the exception of a reconstruction filter (optional) for sending the test signal to the circuit under test (CUT) and the comparator for extracting the digitized waveform from the CUT's response. The periodic waveform generator may comprise a ΣΔ modulator and, optionally, a reconstruction filter between the modulator and CUT. The waveform digitizer may comprise a programmable reference voltage generator for providing a variable voltage reference signal, a comparator for generating a comparison signal from the CUT's analog response signal and the reference signal and means for controlling the reference voltage generator.

An image enhancing system uses a perfect reconstructing filter to process image signals within a specified resolution and enhance the image signals. The image enhancing system includes an image-signal-obtaining unit, a first-signal-group processing unit, a high-pass / low-pass-image obtaining unit, a brightness-adjusting unit, a threshold unit, an amplifying unit, an adder and a normalizing unit. The image-signal-obtaining unit obtains nine image signals from three lines of image signals. The first-signal-group processing unit processes the nine image signals. The high-pass / low-pass-image obtaining unit obtains a high-frequency image signal and a low-frequency image signal. The brightness-adjusting unit adjusts the brightness of the low-frequency image signal. The threshold unit processes the noise of the high-frequency image signal. The amplifying unit amplifies the high-frequency image signal. The adder adds the high-frequency image signal and the low-frequency image signal. The normalizing unit normalizes the result of the adder to obtain an enhanced image.

Sound source separation, without permutation, using convolutional mixing independent component analysis based on a priori knowledge of the target sound source is disclosed. The target sound source can be a human speaker. The reconstruction filters used in the sound source separation take into account the a priori knowledge of the target sound source, such as an estimate the spectra of the target sound source. The filters may be generally constructed based on a speech recognition system. Matching the words of the dictionary of the speech recognition system to a reconstructed signal indicates whether proper separation has occurred. More specifically, the filters may be constructed based on a vector quantization codebook of vectors representing typical sound source patterns. Matching the vectors of the codebook to a reconstructed signal indicates whether proper separation has occurred. The vectors may be linear prediction vectors, among others.

The invention discloses a real-time frequency tracking and harmonic measuring method for the AC sampling of a power system, which is characterized by comprising the steps: an AC sampling input is made to enter an analog-to-digital converter at a fixed sampling rate; the analog-to-digital converter at the fixed sampling rate respectively outputs a first segment of digital signals with the length of N points and a second segment of digital signals with the length of N points to enter a first fast Fourier module; the first fast Fourier module outputs phase angles of the highest frequency components of the first segment of digital signals with the length of N points and the second segment of digital signals with the length of N points, and sends the phase angles into a phase discriminator one after another; the phase discriminator outputs a correction frequency, and sends the correction frequency into a digital signal reconstruction filter; the digital signal reconstruction filter performs digital reconstruction sampling on the first segment of digital output signals with the length of N points, of the analog-to-digital converter at the fixed sampling rate, according to the correction frequency, and sends an obtained N-point reconstruction sampling signal into a second fast Fourier module; the phase angle output of the highest frequency component of the second fast Fourier module is sent into the phase discriminator; the third segment of digital signals with the length of N points, of the analog-to-digital converter, enter the first fast Fourier module; and the phase discriminator judges the phase angle based on accuracy requirements, and outputs a correction frequency to a digital signal reconstruction filter. After the iterations of the process reach M, the output of the second fast Fourier module is the values of respective subharmonic components of the AC sampling signal, and the output frequency of the phase discriminator is the frequency of the AC sampling signal.

The invention discloses a strabismus bias phase center position multi-beam satellite-bone synthetic aperture radar (SAR) imaging method. The method comprises the following steps: corresponding phase compensation is conducted to echo data received by beams of each position by utilizing an azimuth phase compensation function H1, n (f, t); the Doppler frequency spectrum of the echo data after the phase compensation are rebuilt in a distance-Doppler domain by utilizing a position reconstruction filter group P (fa), and Doppler course of a target echo signal is recovered by utilizing a spectrum reduction function H2 (fr, t); and focus imaging is conducted to the echo data by utilizing an improved random multiple access (RMA) algorithm. According to the strabismus bias phase center position multi-beam satellite-bone SAR imaging method, the Doppler course of the echo signal is adjusted through position preprocessing operation, the problem that sampling rate of position signals is insufficient is solved, the Doppler course is rebuilt through strabismus position multi-beam signals, and the problem that position multichannel is not sampled uniformly is solved. Long-distance migration correction and focus imaging are achieved eventually through the RMA algorithm which is based on improved Stolt interpolation operation.

The invention relates to one whole frequency band natural audio direction microphone, which comprises low pass filter, AD converter, signal process system, signal back process system and power amplifier, DA conversion set, re-set filter sets and energy conversion, wherein, the above signal process system comprises more than three frequency division digital filter and more than three filters with inclination and supersonic modulation. The invention can compensate audio signals in whole frequency band range for natural audio effect program to realize whole frequency output.

The invention provides an oversampling 64-time sigma-delta modulation circuit with an effective bit being 18. The circuit comprises an interpolation filter and a sigma-delta modulator. The interpolation filter is used for conducting oversampling interpolation and filtering on digital input signals. The input end of the sigma-delta modulator is connected with the output end of the interpolation filter, the sigma-delta modulator is used for modulating oversampled digital signals, shaping quantized noise introduced by a quantizer, moving noise in the signal bandwidth out of the bandwidth, and meanwhile guaranteeing that transmission of the signals is not affected, and output 1-bit 0 / 1 code streams need to be restored through a backward-stage analog reconstruction filter so as to acquire analog signals. An improved structure is adopted in a half-band filter in the interpolation filter so that the area of the half-band filter can be greatly reduced; for the sigma-delta modulator, a monocycle high-order structure easier to achieve is adopted, the stability problem is analyzed, the high signal to noise ratio is achieved, and meanwhile stability of the modulator is guaranteed.

Said method for control of a reproduction unit (2) for an acoustic field with a number of reproduction elements (31 to 3N) is characterised in comprising: a step for establishing a finite number of coefficients representative of the temporal distribution and in the three spatial dimensions of said acoustic field, a step for determination of representative reconstruction filters for said reproduction unit (2) and at least the spatial configuration of said reproduction unit (2); a step for determination of at least on control signal (SC1 to SCN) for said elements (31 to 3N) by the application of said coefficients to said reconstruction filters and a step for providing said at least one control signal for application to said elements (31 to 3N) for generation of said acoustic field for reproduction.

A method simplifies a data structure representing physical measurements of a real-world phenomena, such as an image of a scene of an object. Input data are first acquired, sensed, or measured. If the data are acquired with a camera, then measurement errors for data points in the data structure are not normally distributed. Therefore, an error term for each data point is constructed according to a Lorentzian estimator. The error term can be determined by taking a difference between the value of the data point, and the value at the data point when a model is fitted to the data structure. The error term is then minimized using a downhill simplex minimization process. Finally, each data point is adjusted by the minimized error term to produce a simplified data structure of the real-world phenomena.

The invention relates to a novel viewpoint rebuilding method in a multi-viewpoint collection / display system of a converge type camera. According to a collection mode of the converge type camera, first, a geometric model of the converge type camera is built, each parameter related to the collection system is comprehensively considered, an optimized condition of distortionless rebuilding viewpoints is obtained through signal sample theory analysis, then spectrum analysis is performed on EPI image on the scene, relation between scene EPI spectral characteristic and scene depth span is obtained. Based on the above analysis, the number of the sample image is determined, and a viewpoint rebuild filter is designed, a sample image is generated for the rebuilt viewpoint through interpolation technology, finally a novel viewpoint of a three-dimension scene in certain range is obtained through the converge model designing and plotting rebuild method. In experiment, verification is respectively performed on an analogue system and a practical system with similar parameters, obtaining excellent rebuilding quality. The invention is of reference value for other camera array types and viewpoint rebuilding system.

The invention discloses a method for constructing a biorthogonal wavelet filter bank. The method includes the following steps: determining the length N of a decomposition low-pass filter h(n) and the length M of a reconstruction low-pass filter (please see the specifications) according to the requirements of wavelet functions to be constructed, wherein the N and the M are even numbers; determining the orders of vanishing moments of dual wavelet generating functions (please see the specifications) to be constructed; calculating the biorthogonal complete reconstruction filter bank according to a formula set to obtain a set of coefficients of the biorthogonal complete reconstruction filter bank; substituting the results into a dual-scale equation to construct a set of biorthogonal wavelet bases (please see the specifications), decomposing a decomposition high-pass filter corresponding to the wavelet function (please see the specifications), and reconstructing a reconstruction high-pass filter corresponding to the wavelet function (please see the specifications). Compared with common db-series wavelet filter banks and common bior-series wavelet filter banks, heart sound signals are processed through the biorthogonal wavelet filter bank constructed with the method, and the better noise reduction effect, the accurate heart sound classified information and the smaller reconstruction error rate can be obtained.

A method for selectively and sequentially converting digital video data signals in three formats, namely, standard definition, progressive scan and high definition formats to analogue form such that the same analogue reconstruction filter can be used for filtering the analogue forms of the signals after conversion. The digital signals prior to being converted are over-sampled at respective over-sampling frequencies for displacing image frequencies to frequencies sufficiently spaced apart from the widest band width signal so that the analogue reconstruction filter can be provided for filtering the widest band width signal as well as the other band width signals and also can be provided with a relatively low roll-off rate of attenuation. The standard definition video data signals are over-sampled at eight times the nyquist sampling frequency, and the progressive scan signals are over-sampled at four times the nyquist sampling frequency, thus displacing image frequencies of the standard definition and progressive scan signals to 108 MHz, so that the centre frequency of the image frequency is displaced to 108 MHz. The high definition video signals are over-sampled at twice the nyquist sampling frequency, thus displacing the image frequency of the high definition signal to 148.5 MHz, so that its centre frequency is displaced to 148.5 MHz. This, thus, leaves a frequency band gap of approximately 66 MHz between the high definition band width signal and the closest image frequency, namely, that of the progressive scan signal. The method can also be applied to the conversion of analogue video signals to digital form whereby the same analogue anti-aliasing filter can be used for the analogue signals in the three formats.

The present disclosure is directed to an augmented reality surgical system for viewing an augmented image of a region of interest during a surgical procedure. The system includes an image capture device that captures an image of the region of interest. A controller receives the image and applies at least one image processing filter to the image. The image processing filter includes a spatial decomposition filter that decomposes the image into spatial frequency bands. A temporal filter is applied to the spatial frequency bands to generate temporally filtered bands. An adder adds each band spatial frequency band to a corresponding temporally filtered band to generate augmented bands. A reconstruction filter generates an augmented image by collapsing the augmented bands. A display displays the augmented image to a user.

A method for converting multi-bit digital video data signals to analogue form wherein the video data signals are in three formats, namely, standard definition, progressive scan and high definition formats, which permits a relatively low order analogue reconstruction filter to be used for filtering the analogue form of the signal. The three formats of the digital video data signals are over-sampled at relatively low over-sampling rates, the standard definition video signals being over-sampled at eight times the nyquist sampling frequency, the progressive scan format signals being over-sampled at four times the nyquist sampling frequency, while the high definition format signals are over-sampled at twice the nyquist sampling frequency. By virtue of the fact that the Video data signals are over-sampled at relatively low over-sampling rates, the over-sampled signals are subjected to sigma-delta modulation in a relatively low order sigma-delta modulator, namely, a second order sigma-delta modulator for noise shaping the quantisation noise. This, thus, permits the analogue reconstruction filter to be provided as a third order filter, and thus of relatively low order.

An original image is up-sampled to a final image by constructing at least two sub-banded filtered images of the original image with 2D wavelet-based decomposition filters, each filtered image being of the same resolution as the original image. Each sub-banded filtered image is then mapped into a larger filtered image of the same size as the final image. Pixels in each larger filtered image which were not mapped in from the sub-banded filtered images are interpolated or left blank. The larger filtered images are then filtered with 2D reconstruction filters and combined to form the final up-sampled image. The invention has the advantage that high quality up-sampled images can be created in real time suitable for high definition video.

A pair of M-path perfect reconstruction filter banks, one for analysis and one for synthesis avoid aliasing of channel filter band-edges by operating at an output sample rate of 2-samples per channel bandwidth. The non-maximally decimated analysis filter performs an M / 2-to-1 down sample operation to obtain the 2-samples per channel output rate. The corresponding synthesis bank accepts signal samples at 2-samples per symbol and up-samples 1-to-M / 2 to obtain the original and desired output sample rate matched to the input sample rate. Changes in composite system bandwidth are achieved by enabling or disabling the connection between the analysis filter bank and the corresponding synthesis filter bank.

The invention is suitable for the technical field of computers, and provides a deep neural network compression method and system, and terminal equipment, and the method comprises the steps: inputtingtest sample data, obtaining an original feature map of an Lth layer of a deep neural network, and determining a redundant filter of the Lth layer according to the original feature map of the Lth layer; pruning the Lth layer according to a redundant filter; obtaining an original feature map of the (L + 1) th layer and a feature map of the L-th layer after pruning; inputting the original feature mapof the (L + 1) th layer and the pruned feature map of the Lth layer into a filter learning model, and automatically learning and outputting a reconstruction filter of the (L + 1) th layer through thefilter learning model; inputting the pruned feature map of the L th layer into a reconstruction filter of the L + 1 layer to obtain a target feature map of the (L + 1) th layer, according to the method, carrying out pruning and reconstruction based on the feature map, the filter is automatically learned and reconstructed in combination with the influence of pruning, the classification accuracy of the compressed deep neural network model is ensured while the structural sparsity of the filter is realized, and the calculation efficiency is improved.

A method filters a depth image, wherein each depth image includes an array of pixels at locations (x, y), and wherein each pixel has a depth. A moving window is applied to the pixels in the depth image, wherein a size of the window covers a set of pixels centered at each pixel. A single representative depth from the set of pixel in the window is assigned to the pixel to produce a processed depth image. Then, each pixel in the processed depth image is filtered to correct outlier depths without blurring depth discontinuities to produce a filtered depth image.

The invention relates to an ultra wide band transmitter and a design method thereof. The transmitter uses current as a carrier of signals, adopts a zero intermediate frequency structure, and consists of modules such as an IQ two-path current top-shaped digital-to-analog converter with high sampling rate, an IQ two-path current type low-pass reconstruction filter, an IQ two-path current type upperfrequency mixer, a controllable gain radio-frequency amplifier, and the like. The digital-to-analog converter converts signals from a digital baseband into corresponding current signals, then analog baseband signals are reconstructed from the current signals by the low-pass filter, and the signals directly are sent to the current type frequency mixer to finish modulation and upper frequency mixing and finally are amplified by the radio-frequency amplifier and transmitted to the outside by antennas; the transmitter simplifies the structure and is beneficial for improving the linearity degree of the system and reducing the power consumption. The transmitter is designed by adopting an on-chip integrated digital-to-analog converter, and has great advantage in low voltage and high frequency design owning to the collaborative design among the modules.

A system and method is provided for efficient analog reconstruction of a digital signal to an intermediate frequency (IF). The digital signal can be a complex baseband digital signal that is sourced from a near perfect reconstruction filterbank (NPRFB). The output of the NPRFB, after some manipulation in the digital domain, is converted to the analog domain at IF. The IF analog signal can then be modulated with a high frequency RF signal or further mixed with a RF transmission frequency signal for RF transmission.

The invention discloses a broadband high-dynamic radio-frequency direct quadrature modulation device and relates to a direct quadrature signal modulation device having a wide working band and a wide output dynamic scope and controlled by a program in the field of satellite navigation test. The broadband high-dynamic radio-frequency direct quadrature modulation device is mainly composed of a baseband signal input and control interface, a difference reconstruction filter, a direct quadrature modulation module, a local oscillator signal generating module, a communication and control module, a large-step damping module, a small-step damping module, a power supply terminal, a program burn interface, and the like. The broadband high-dynamic radio-frequency direct quadrature modulation device realizes the baseband quadrature signal direct quadrature modulation up-conversion with wide working band and wide output dynamic scope by adopting a high-performance broadband phase-locked loop technology and a large-and-small step combined damping control technology, has the characteristics of miniaturization, simple interface, convenience in debugging, perfect function, excellent performance, flexibility, controllability, and the like, and is especially suitable for a Beidou navigation system as a radio-frequency emission front-end of a multi-system navigation signal or navigation interference signal generating device.