121 results about "Zero filling" patented technology

The primary synchronization signal detection method in an LTE system. The method comprises the following steps: converting cross-correlation process to linear convolution, converting the linear convolution into circular convolution in a zero filling mode, and performing Fourier transform FFT operation to obtain a primary synchronization signal PSS sequence dmf(n) in a local frequency domain; performing the Fourier transform FFT operation on a received signal sequence, and performing point-to-point multiplication on the received signal sequence and the local primary synchronization signal PSS sequence to obtain a cross-correlation power sequence, judging whether the primary synchronization signal PSS is successfully retrieved by judging the peak to average ratio of the cross-correlation power sequence, and if so, calculating a PSS signal position according to the peak position, and calculating a call intra-group identifier; and otherwise, indicating that the retrieval is failed. The primary synchronization signal detection method has the beneficial effects of reducing the complexity of the cross-correlation operation while keeping the wide adaptability of the cross-correlation operation in PSS synchronization, improving the operation accuracy while reducing the single-step operation data size and improving the operation efficiency and accuracy.

The invention relates to the field of noise, in particular to a sound field separation and reconstruction method applicable to sound field separation and reconstruction of large sound sources. The sound field separation and reconstruction method comprises the following steps of: acquiring sound pressure and normal particle vibration velocity on a measuring surface; performing zero filling extension on the measuring surface positioned between two sound sources; acquiring transfer matrixes between the extended measuring surface and surfaces of the two sound sources, namely sound source faces; building a transitive relation between the sound pressure and the normal particle vibration velocity on the measuring surface; and acquiring the sound pressure and the normal particle vibration velocity of the first sound source face and the sound pressure and the normal particle vibration velocity of the second sound source face. A sound field is separated and reconstructed by using the single-measurement-surface and local near-field acoustical holography method. The sound field separation and reconstruction method has the characteristics of simplicity, short calculation time and high calculation efficiency, and can be widely applied to the measurement of near-field acoustical holography, the measurement of material reflecting coefficients, the separation of scattering sound fields and the like of large-size sound-source sound fields.

A multi-color off-axis digital holographic system and the imaging method thereof are disclosed. The multi-color off-axis digital holographic system comprises: a plurality of light emitting diodes, for provide a red (R) beam, a green (G) beam and a blue (B) beam; an interference object lens module, for receiving the R, G, and B beams to generate a beam containing an interference signal; a color imaging device, for receiving the beam containing the interference signal and thus forming a hologram on a surface of the color imaging device by holographic interference while registering the hologram; and a processing device, for receiving the registered hologram form the color imaging device; wherein the processing device perform a zero-filling and reconstructing operations upon the received hologram to obtain phase information of the R, G and B beams. With the aforesaid system, a three-dimensional surface profile with respect to a RG synthetic wavelength is obtained according to a calculation using the phase information of the R and G beams as well as the wavelengths thereof, and similarly, a three-dimensional surface profile with respect to a GB synthetic wavelength is obtained according to a calculation using the phase information of the G and B beams as well as the wavelengths thereof. Thereafter, by performing a calculation using the RG synthetic wavelength and its phase as well as the GB synthetic wavelength and its phase, an overall three-dimensional surface profile with respect to a complete synthetic wavelength is obtained.

The invention relates to a harmonic energy metering method of Nuttall self-convolution weighted Fourier transform. The method comprises the following steps that: a discrete Nuttall window sequence with a length of N is constructed; convolution operation is carried out on the discrete Nuttall window so as to obtain a convolution sequence; zero filling is carried out on the head or the tail of the convolution sequence so as to obtain a discrete Nuttall self-convolution window; the Nuttall self-convolution window sequence is utilized to carry out weighted operation on a discrete power signal; Fourier transform is employed to obtain a discrete frequency spectrum of the power signal; spectral lines with a greatest amplitude and a second great amplitude are searched nearby a harmonic frequency corresponded to the discrete frequency spectrum; fitting is carried out on a peak value spectral line by utilizing a least square method so as to obtain an amplitude, a frequency and an initial phase angle of a power harmonic wave, so that integral harmonic energy metering in a complex power signal is realized.

In order to solve the problem that a traditional machine learning algorithm has difficulty in feature extraction and confusion of similar actions to classification and recognition, the invention provides a multi-size input HAR algorithm based on an improved LSTM-CNN. According to the algorithm, an optimized CNN convolutional neural network and a multi-layer bidirectional dynamic LSTM long-short time memory network are fused to identify HAR human body behavior activities. The method comprises multi-size input and model classification and recognition, the multi-size inpu subjects the data to theoperations of noise reduction, mean value-variance normalization, zero filling and the like to generate two different dimensions of data as model input, and an optimized CNN network and a multi-layerbidirectional dynamic LSTM network are fused for classification and recognition through model classification and recognition. According to the method, the optimized convolutional neural network and the multi-layer bidirectional dynamic long-short-term memory network are fused to construct the classifier, so that the method has good expansibility and robustness, and can realize high-precision human behavior activity recognition.

The invention discloses a method for reconstructing sparse MRI (Magnetic resonance imaging) based on a convolutional neural network in combination with an iterative method. The method comprises the steps of firstly, preparing a data set which includes training data and test data, wherein the training data are used for training the network, and the test data are used for testing the trained data, each group of data includes a group of samples and labels, the samples are low-quality high-frequency images and low-frequency images with noise and image artifacts obtained by respectively zero-filling and reconstructing low-frequency data and high-frequency data which are acquired by dividing height down-sampling k-space data, and the labels are high-quality MR images without noise or image artifacts corresponding to the low-quality images; and training two networks with identical structures by use of the low-frequency data and the high-frequency data, respectively, wherein one network is used for reconstructing high-frequency k-space data, the other network is used for reconstructing low-frequency k-space data, and the addition of two reconstruction results is a final needed reconstruction result. According to the method for reconstructing the sparse MRI based on the convolutional neural network in combination with the iterative method, less k-space data are utilized, the reconstruction speed is faster and the image quality is higher.

The invention discloses a GNSS new-system signal capturing device and method. The device comprises a difference correlator module, a carrier wave NCO generator, a local PN code generator, a correlator, a frequency mixer, a segment correlation zero-fill FFT module and a capture detection control module. The difference correlator module determines the PN code phase offset of a received satellite signal via a digital intermediate-frequency input signal, a difference output signal and a local carrier wave are mixed and demodulated to the base band, an in-phase branch I and an orthogonal branch Q are obtained via the correlator and the local PN code generator, the segment correlation zero-fill FFT module carries out coherent integration and segment noncoherent accumulation on operation results, and after zero-filling of the R-point accumulation result, S-point FFT is carried out to complete rapidly capture the GNSS new-system signals. The PN code phase offset of the received satellite signal can be rapidly determined, the Doppler frequency estimation precision can be improved via segment correlation zero-fill FFT, the device is simple, the false capture rate is low, and new-system signals can be rapidly captured.

The invention discloses a method for SPEED rapid magnetic resonance imaging based on wavelet domain sparse representation. The method for SPEED rapid magnetic resonance imaging based on wavelet domain sparse representation comprises two steps of regular under-sampled data collection and SPEED image reconstruction based on wavelet domain sparse representation. The regular under-sampled data collection comprises two steps of k-space regular under-sampling and k-space central domain complete sampling. The SPEED image reconstruction based on wavelet domain sparse representation comprises seven steps of zero filling Fourier reconstruction, discrete wavelet transformation, wavelet domain double-layer sparse ghost model construction, overlapped ghost separation based on a least square error method, discrete wavelet reverse transformation, rectification summation of a plurality of ghost images and image reconstruction. The discrete wavelet transformation is used for SPEED data sparse representation, the sparseness of the image is improved through discrete wavelet transformation, and further the quality of the SPEED reconstruction image is improved.

The invention discloses a fixed-point multiplying accumulator, which comprises a Booth coding unit, a left shift logic unit, a Booth decoding / partial product generation unit, a compression tree, a compressor and a summator, wherein the Booth coding unit is used for coding a multiplier; the left shift logic unit is used for enabling a multiplicand to leftwards shift for one bit to carry out lower-bit zero filling when the multiplicand is a decimal; the Booth decoding / partial product generation unit is used for generating partial products; the compression tree is used for compressing the partial products and outputting two piece of compressed data; the compressor is used for compressing data provided by the accumulator and the compression tree; and the summator is used for carrying out an addition operation on the data output by the compressor to output a result. By use of the fixed-point multiplying accumulator, fixed point integer / decimal multiplication, multiply-accumulation / subtraction functions and overflow judgment functions are realized, the time delay of a multiply-accumulator is not increased, and the performance of a processor is guaranteed.

The invention discloses a data compression method and device and a data transmission method and system. The data compression method comprises the steps that amplitude information and phase information of all IQ data pairs in each IQ data pair group are read respectively; the phase information of all the IQ data pairs is subjected to bit cutting to obtain corresponding effective bit information; the amplitude information and the effective bit information in each IQ data pair group form new IQ data pairs, and the new IQ data pairs are compressed. According to the data compression method and device, in the compression process, the phase information is subjected to bit cutting before compression is conducted, so that the information compression amount is reduced. According to the data transmission method and system, in the compression process, the phase information is subjected to bit cutting before compression is conducted; in the decompression process, the effective bit information is subjected to zero filling to obtain the phase information, and finally the phase information is converted into IQ data, so that the information transmission amount is reduced.

The invention provides an array beamforming method by quickly expanding and dragging a broadband frequency domain. The method combines ETAM passive synthetic aperture technology and a broadband frequency domain beamforming method, and comprises the following steps of: 1) receiving space signals with a linear array to acquire time domain signals of M array elements; and taking number of overlappedarray elements as N; 2) carrying out fast Fourier transform on k snapshot and k+1 snapshot data in a time domain; 3) determining frequency band range of target radiation signals, and calculating phase shift parameters according to the overlapped array elements of any two adjacent snapshots; 4) carrying out phase compensation on different frequency components respectively to acquire virtual synthesized array elements; 5) repeating the steps, and carrying out K times of aperture synthesis; 6) carrying out FFT transform on zero filling of the synthesized array element signals in a spatial domain, and then transferring zero frequency components of the acquired data to a spectrum center; 7) calibrating each frequency point to acquire beam output of different frequency points; and 8) synthesizing beamforming results of different frequencies so as to realize accurate positioning of targets.

The invention discloses a correlation function-based moving-target parameter estimation method. The method comprises steps of 1: reading in original moving target echo data and related imaging parameters; 2: performing Fourier transform processing for azimuth; 3: performing compensation by multiplying azimuth and CS factors; 4: performing Fourier transform processing for range distance; 5: multiplying the range distance and distance compensation factors so as to process distance compressing processing; 6: performing Fourier inversion processing for range distance; 7: performing correlation function processing for distance-Doppler domain; 8: performing Fourier inversion processing for azimuth; 9: performing zero filling for frequency domain, and adding sampling in time domain so as to perform interpolation processing or related processing results and get the maximum value; and 10: estimating target speed via the related maximum value. According to the invention, estimation of target azimuth and radial speed are simultaneously achieved, precision for estimation speed of moving targets of a current satellite born SAR is increased, and a novel detection method is provided.

The invention discloses a high-precision method for acquiring an FMCW radar distance measurement value and aims to discloses a distance measurement value method with high ranging precision and capableof reducing frequency spectrum errors. The method is implemented by adopting the following technical scheme; sending target beat signals subjected to analog-digital conversion into four parallel timedomain extraction module; sending digital beat signals and sampling rate data acquired respectively into a data zero fill module for time domain zero filling; acquiring discrete frequency domain signals having a length of Ns through discrete Fourier transform (DFT) on time domain signals having a length of Ns acquired by the four time domain extraction module through corresponding DFT modules; sending the discrete frequency domain signals are sent into a data interception and splicing module by the four DFT modules in common-end manner for frequency domain data splicing operation, and an FMCWradar distance measurement value is acquired through a distance measurement value acquiring module based on a mathematic interpolation operation result.

The invention discloses a wavelet-denoising-based SC-FDE channel estimation method in a short wave channel. The wavelet-denoising-based SC-FDE channel estimation method includes the steps of firstly, partitioning sent data, periodically inserting pilot frequency between data blocks, extracting information at the pilot frequency positions at a receiving end, and estimating channel frequency responses at the pilot frequency with an LS algorithm; secondly, carrying out IDFT on obtained estimation values to be in a time domain, factoring the estimation values in a wavelet domain according to a multi-Mallat algorithm, filtering out noise with a threshold denoising method, and reconstructing noise-eliminated signals; finally, carrying out interpolation zero filling on the reconstructed signals, carrying out DFT on the reconstructed signals to be in a frequency domain, accordingly obtaining all the channel frequency responses, and completing channel estimation. According to the wavelet-denoising-based SC-FDE channel estimation method in the short wave channel, the LS estimation left noise is eliminated with a wavelet denoising method, influences of the noise are further suppressed accordingly, the system bit error rate is lowered, and the system performance is improved.

The invention relates to a transmission method for wireless local area network signaling. The method comprises the following steps of: performing primary binary convolutional coding (BCC) on a very high throughput-signal-B (VHT-SIG-B) bit sequence; performing repeating and zero filling operation on a generated coded bit sequence to form a VHT-SIG-B coded bit sequence; and before space mapping, performing transmission processing of only one 80MHz sub-band for a 160MHz channel and an 80+80MHz channel, and performing space mapping and subsequent processing on an output signal which is taken as signals of two 80MHz sub-bands. A transmission device for the wireless local area network signaling comprises a coding module, a data stream processing module and a transmission module. The repetitiveness of a VHT-SIG-B extended bit sequence and the same part in the transmission processes of different sub-bands of the 160MHz channel and the 80+80MHz channel are fully utilized, so that repeated operation is avoided, and processing complexity is remarkably lowered.

The invention discloses a character compression encrypting method, belonging to the field of information security protection. The character compression encrypting method comprises the following steps: converting characters into region-position codes corresponding to a Unicode character set, and converting the region-position codes into binary codes; then carrying out zero filling operation on the binary codes; and finally converting the binary codes into new characters, and simultaneously carrying out elliptical encryption on information to be encrypted firstly. The method adopted by the invention is simple to realize, is capable of carrying out the safety encryption on Chinese characters, letters, figures, and the like simultaneously, and is easy to implement; the character compression encrypting method has better anti-attack performance than other existing Chinese character encryption algorithms, and the information volume after encryption is smaller; therefore, the lossless restitution of ciphertext can be realized.

The invention relates to a carrier synchronization method in a UFMC system, comprising the following steps: first, generating two sets of orthogonal codes at a sender, inserting the orthogonal codes as pilot sequences into original data, and modulating, filtering and sending the original data through UFMC; then, performing time-domain zero filling and 2N-FFT transform at a receiver, extracting data on even bands, sending the data to a filter matching the sender, getting the estimated value of data information using a zero forcing equalizer, extracting pilot signals and calculating the correlation between the two sets of pilot signals, and minimizing the correlation to achieve UFMC system carrier frequency synchronization; and finally, using an iterative algorithm to improve the accuracy of frequency offset estimation. According to the invention, the carrier frequency offset of the system is estimated by inserting a pilot frequency into each sub band of UFMC, so good properties of the UFMC system are maintained, and the accuracy and stability of frequency offset estimation are improved through iteration. Therefore, the bit error performance of the system is enhanced, the quality of communication is improved, and the design complexity of the filter in the UFMC system is reduced to a certain extent.

The invention discloses an azimuthal multi-channel FMCW SAR sliding bunching imaging method. The method comprises: frequency-modulation continuous wave signals are emitted by a channel and the frequency-modulation continuous wave signals are received simultaneously by azimuthal M channels to obtain-path azimuthal under-sampling frequency-modulation continuous wave sliding bunching SAR echo signals; a Deramping operation is carried out on each path of echo data at an azimuth to eliminate a change of a Doppler center with time; azimuth reconstruction is carried out on the M-path echo data after the Deramping operation to obtain one path of echo data; frequency domain zero filling and Ramping operation are carried out on the echo data after azimuth reconstruction to obtain a non-aliasing scene echo signal; and imaging processing is carried out on the obtained non-aliasing scene echo signal to obtain a focused SAR image.

The invention discloses a Winograd convolution splitting method for a convolutional neural network accelerator. The method comprises the following steps: 1) reading an input and a convolution kernel of any size from a cache of the convolutional neural network accelerator; 2) judging whether convolution splitting is carried out or not according to the convolution kernel size and the input size, andif convolution splitting needs to be carried out, carrying out the next step; 3) splitting the convolution kernel according to the size and the step length of the convolution kernel, and splitting the input according to the size and the step length of the input; 4) combining and zero-filling the split elements according to the size of the convolution kernel, and combining and zero-filling the split elements according to the input size; 5) performing Winograd convolution on each pair of split input and convolution kernels; 6) accumulating the Winograd convolution results of each input and convolution kernel, and 7) storing the accumulation results in a cache of the convolutional neural network accelerator, so that the convolutional neural network accelerator can support convolution of various different shapes by adopting one Winograd acceleration unit.

The invention discloses a sub-band splicing method based on internal calibration data. The sub-band splicing method comprises the following steps: step I: extracting in-band errors and time delay between channels; step II: processing amplitude errors and time delay extracted by an internal calibration signal in each sub-band; step III: shifting internal calibration compression positions of different sub-bands to a zero moment, and processing phase errors of each sub-band; step IV: reading radar echo data of sub-bands, and performing in-band error compensation of a single sub-band; step V: compensating in-band errors of each sub-band signal, and calculating time delay of each sub-band; step VI: performing upsampling on a frequency domain signal of each sub-band in the step V in a tail zero-filling mode; step VII: performing splicing process in a frequency domain; step VIII: obtaining a spliced time domain signal. The sub-band splicing method disclosed by the invention not only can compensate amplitude phase errors in the sub-bands, but also can compensate amplitude phase errors among the sub-bands as well as time delay among different sub-bands, and realizes coherent fusion processing of different sub-band signals.

The invention provides a signal spectrum zooming method based on linear frequency modulation z transform. A symmetrical linear system h(n) is obtained by changing the expression form of a Bluestein equation and carrying out zero filling and periodic extending on the linear system h(n); the method for calculating the Fourier transform of an L point linear system sequence h(n) through a stored L / 2+1 point linear system sequence is put forward; the result q(n) of the circular convolution is horizontally moved leftwards by N0 units, the front M values obtained after horizontal movement are multiplied with k (k=0,1,..., M-1), and then the signal partial spectrum with the improved frequency resolution can be obtained. By means of the scheme, the storage data size of the linear system sequence and the Fourier transform of the linear system sequence can be reduced, the storage space is saved, the Fourier transform of an L point linear system can be calculated through the stored L / 2+1 point linear system data, the calculated quantity of an FFT is reduced, and the calculating efficiency is improved.

The invention provides a construction and coding method for multi-code rate and multi-code length LDPC (Low Density Parity Check) codes. The method comprises the steps of constructing a template matrix; constructing check matrixes according to the template matrix; receiving an information bit length, a coding bit length and input information bits; selecting a target template matrix, acquiring a coding check matrix according to the target template matrix, acquiring a shortened check matrix according to the coding check matrix, acquiring an intercepted check matrix according to the shortened check matrix, and carrying out zero filling on the input information bits to acquire filled information bits; carrying out LDPC coding on the filled information bits to acquire coding bits; removing zerofilled bits and partial check bits in the coding bits, and deleting partial information bits contained in the coding bits to acquire output coding bits. According to the method, the multi-code rate and multi-code length QC-LDPC (Quasi-Cyclic LDPC) codes of which information bit and check bit lengths are changed bit by bit can be constructed, moreover, the performance of the QC-LDPC codes is guaranteed, and the design and realization of a coding modulation system based on the LDPC codes are simplified.

The invention discloses a divider and a division processing method. The method comprises unsigned and signed division operations. An unsigned divider comprises a zero judgement unit, a zero filling unit, a shift unit, a judgement processing unit and a result processing unit. The zero judgement unit is used for judging whether a divisor is zero or not. The zero filling unit is used for filling zero to a dividend according to decimal digits to be reserved by a quotient. The shift unit is used for performing leftward movement operation of the divisor. The judgement processing unit is used for cyclically judging according to the leftward movement operation of the shift unit and the zero filled dividend output by the zero filling unit, determining each member of the quotient and inputting to the result processing unit. In the method disclosed by the invention, each member of the quotient can be obtained in turn through the approximation bit by bit by using the cycle shift of the divisor; the decimal digits to be reserved by the division result can be set according to requirements; the complex iterative operation can be avoided through the approximation bit by bit; the delay is lower; the whole operation process completely depends on shift, substraction and comparison; and the portability is strong.

A single-scanning quantitative magnetic resonance T2 * imaging method based on residual network reconstruction relates to a magnetic resonance imaging method. Four small-angle excitation pulses with the same deflection angle are utilized, a period of evolution time is formed after each excitation pulse, so that the transverse relaxation time of each echo signal is different. A shifting gradient ofa frequency dimension and a phase dimension is added after each excitation pulse, so that the signals generated by different excitation pulses are different in the position of the k space. In this way, a plurality of gradient echo signals with different transverse relaxation time are obtained in one-time sampling. The sampling signals are subjected to normalization, zero filling and fast fouriertransformation and then input to a trained residual network to be reconstructed to obtain a quantitative T2 * image. The training data of the residual network are derived from simulation data, and theinput image of a network is obtained via a random generation template through simulating the experiment environment for sampling. The template serves as a label, and a mapping relation between an input image and an output image is obtained through training.

The invention discloses a reference waveform selection method based on the joint of primary and secondary envelopes in a cross-correlation method-based gas ultrasonic flowmeter. For the problem of 5 [mu]s layering of a transit time calculation value due to use of a traditional reference waveform when the gas flow speed in a pipeline is relatively high, an improved reference waveform selection method is provided. According to the method, a zero crossing point before the inverse fourth wave peak of the primary envelope of an average waveform is used as a starting point; a zero crossing point after the fourth wave peak of the secondary envelope is used as a cutoff point; a waveform between the starting point and the cutoff point is determined as a standard waveform; the rear end of the standard waveform is filled with zero, so that the number of signal points after zero filling is the same as that of collected signal sampling points when the ultrasonic flowmeter works; and a signal waveform after the zero filling of the standard waveform is determined as a reference waveform. According to the method, the problem of the layering of the transit time calculation value is solved, and themeasurement precision and the repeatability of the gas ultrasonic flowmeter are improved.

The invention discloses an image filling method and a device in FPGA acceleration of a convolution neural network. The method comprises steps that an image to be processed is acquired; The image to beprocessed is displayed in the form of a first matrix, and each matrix element in the first matrix is one pixel data of the image to be processed. Performing a zero filling process on the first matrixto obtain a second matrix; The first column and the last column of the second matrix are zeros, and each matrix element in the other columns is one pixel data of the image to be processed; Alternatively, the first row and the last row of the second matrix are zero, and each matrix element in the other row is one pixel data of the image to be processed; Performing convolution calculation accordingto the second matrix and convolution weights to obtain a third matrix; The size of the third matrix is the same as the size of the first matrix. In this way, the calculated image can have the same size as the original image.

The invention discloses a real-time internal calibration method and device for a multi-channel digital receiver. The method comprises the following steps: carrying out matched filtering and frequencyspectrum zero filling on internal calibration data in all channels to obtain signal impact response data of all channels; obtaining amplitudes, phases, and time-delay inconsistency characteristics among all channels based on the impact response data of all channels; carrying out amplitude and phase inconsistency characteristic correction on echo collection data of all channels to obtain correctedecho signals; and carrying out time-delay inconsistency characteristic correction on the corrected echo signals. Therefore, a problem of difficult realization of real-time internal calibration correction for amplitude, phase, and time delay inconsistency characteristics among all channels by the multi-channel radar system is solved; delay correction is realized by time-domain filtering; and the computational complexity is reduced substantially.

The invention discloses an improved Quinn algorithm based Doppler frequency estimation method. Before the Quinn algorithm is used to perform Doppler frequency estimation, the Fourier transform FFT operation for zero filling is performed on the signal, so that the number of spectral lines in a signal spectrum main lobe is increased, and more spectral line information is used to correct the peak position, so that a more accurate Doppler frequency estimation value can be obtained. The method comprises the steps of: performing double zero-filling on a spread spectrum signal x(n) of a N-point, andperforming an FFT operation with a point number of 2N on the double zero-filled signal to obtain a frequency domain signal X(k); selecting a peak value |X(k0)| of the frequency domain signal X(k) amplitude, wherein k0 is a frequency domain position coordinate corresponding to the peak |X(k0)|; using the Quinn algorithm to perform frequency domain interpolation, and determining the interpolation direction r of the Quinn algorithm, where r= +1 indicates that the interpolation direction is going to the right, and r=-1 indicates that the interpolation direction is going to the left; and calculating the movement amount of the spectral line interpolation movement of the Quinn algorithm (shown in the description); and calculating the Doppler frequency estimated value (shown in the description), wherein delta f is the discriminating interval of the FFT operation.

The invention relates to a blind estimation method of rotation angles of digital images. If rotary gray level images are color images, luminance components are obtained after spatial alternation of YCbCr, difference images are obtained by conducting second order difference or laplace filtering on images to be detected, and the difference images are processed line by line (or row by row); a vector quantity V with the length of N0 is obtained after zero filling at the tail end, and the N0 is integer power of 2; the discrete Fourier transform of the V is obtained, the magnitude spectra D=|DFT(V)| is obtained; the D is searched, if the D(k) is a maximum value in a neighborhood region [k-delta, k+delta], and a peak value counter c(k)=c(k)+1. Normalization is carried out on the c(k) to obtain a frequency-peak value counting picture, and the rotation angles of the images can be determined by detecting the frequency corresponding to the peak value. Blind estimation can be achieved without priori knowledge such as original images before rotation or reference templates, the blind estimation method is basically not influenced by image joint photographic expert group (JPEG) compression storage quality factors, the algorithm complexity is low, detection precision is high, size of the images is big, and estimated accuracy is high.

The invention discloses frequency bias estimation equipment, a method and a communication device. The frequency bias estimation equipment includes a path selection module, a pre-processing module, a zero filling module, a spectrum analysis module and an interpolation module. The path selection module is used for executing channel estimation to the input signal in order to selection a plurality of paths. The pre-processing module is used for multiplying a combination signal with complex conjugate of the input signal, wherein the combination signal is obtained by rake combining the channel estimation of the selected path with input signal execution class. The zero filling module is used for filling zero on the tail of the pre-processing signal. The spectrum analysis module is used for performing spectrum analysis to the zero filled signal in order to search a plurality of peaks. The interpolation module is used for estimating frequency bias between the sending part and the receiving part by interpolation with the plurality of peaks.