5 results about "Circulant matrix" patented technology

The invention discloses a weak cyclostationary signal detection method based on an unmanned aerial vehicle-mounted passive radar network, and the method comprises the steps: firstly building an active radar transmitting signal and unmanned aerial vehicle passive radar receiving signal model, and then building a binary hypothesis model based on cyclostationary generalized likelihood ratio detection through block cyclic matrix approximation; and constructing generalized likelihood ratio test detection statistics and obtaining local detection results of the unmanned aerial vehicles under a colored noise background by using the cyclostationary characteristic of target echoes, finally fusing the local detection results of the unmanned aerial vehicles based on a Bayesian minimum risk criterion at a fusion center, and outputting global detection judgment. And a detection process of cyclostationary generalized likelihood ratio test-Bayesian minimum risk fusion is formed. According to the method, the detection statistics are constructed by using the cyclostationary characteristics of the target echoes, the global detection accuracy and robustness can be improved under the conditions of low signal-to-noise ratio and complex colored noise, and the method is suitable for a multi-unmanned aerial vehicle cooperative passive radar detection scene.

Certain example embodiments may relate to a 5G and / or 6G communication system for supporting higher data transmission rates than 4G communication systems such as LTE. A method performed by a transmitter in a communication system, which method may include: determining the number of input bits; determining a base matrix on the basis of the number of input bits; determining a lifting size (Z) on the basis of at least one of the number of input bits or the base matrix; determining a parity check matrix on the basis of at least one of the base matrix or the lifting size (Z); and performing encoding on the basis of the parity check matrix and the input bits. The size of the parity check matrix is Z×(k+1) Z, the last column block of the parity check matrix corresponds to parity bits, a Z×kZ column block of the parity check matrix corresponds to information word bits, all of modulo-Z values of differences in cyclic shift values corresponding to circulant permutation matrices constituting a Z×Z circulant matrix are different, and all of the circulant matrices corresponding to the information word bits are composed of three or more circulant permutation matrices.

The application discloses a three-dimensional neural network processing method, an image processing method, an image processing system and a storage medium, and belongs to the technical fields of neural networks and image processing. The three-dimensional neural network processing method based on a block circulant matrix model overcomes the irregular memory and calculation problems existing in the previous pruning and other model compression methods by constructing a block circulant matrix model, calculating an acceleration model and a full frequency domain model, and proposes to compress a three-dimensional neural network 3D CNN by using the block circulant matrix, and further utilizes fast Fourier transform FFT to accelerate calculation, so that the storage and calculation compression effects are significantly achieved on the premise of keeping the model structure regular. On this basis, the activation, batch normalization and pooling operations in the frequency domain are introduced, the frequent time domain / frequency domain switching overhead caused by the fast Fourier transform FFT is further eliminated, so that full frequency domain calculation is realized, and the calculation overhead during 3D CNN model reasoning is further reduced.

In an automotive radar system, a measurement vector is determined using signals received a plurality of radar receiver modules. An expression is determined that defining an iteration of an optimization problem configured to determine an optimized output amplitude vector based on the measurement vector, wherein the expression includes a first parameter that is a hermitian-centrohermitian matrix or a circulant matrix. The automotive radar system includes a neural network and each node of the neural network solves iterations of the expression to determine an optimized value of the first parameter. A final node of the neural network determines the optimized output amplitude vector based on the optimized value of the first parameter and an estimated direction of arrival of a first object is determined using the optimized output amplitude vector.

The invention provides a CDVFT fine tuning method based on cyclic matrix inversion. The CDVFT fine tuning method is efficient in memory. The CDVFT fine tuning method comprises the following steps: acquiring an input column vector and a weight increment; performing forward processing based on the input column vector and the weight increment to obtain an intermediate calculation result and forward output, and only storing the forward output in the memory in the model training stage; during the period of back propagation during training, carrying out inverse operation on forward propagation on forward output to obtain a tensor required by gradient calculation; carrying out inverse operation on element-by-element multiplication of a product of an intermediate state vector and a diagonal matrix vector by using an element-by-element division method; the calculation of the intermediate state vector and the cyclic matrix vector is based on inverse operation of cyclic matrix inversion. According to the method, the tensor stored in the forward propagation period of fine tuning is reduced by using the characteristic that the cyclic matrix can be used for rapid inversion, and the tensor is recovered in the reverse propagation period, so that the peak memory occupation in the fine tuning process is reduced.