32results about How to "Noise robustness" patented technology

The invention discloses a 3D terrain imaging system of the interferometric synthetic aperture radar (InSAR) and an elevation mapping method thereof, which mainly solve the problems that the existing InSAR has bad imaging pragmaticality and can not implement 3D elevation mapping on the fast-changing terrain and the transilient terrain. The system comprises three sub-aperture antennas, a radar transmitter, a radar receiver and an imaging data processor; the imaging signal processor comprises a SAR image processing unit and an InSAR image processing unit. The invention receives radar echo through the three sub-apertures, then conducts SAR imaging process on the radar echo respectively received by the three sub-apertures, and then conducts InSAR imaging process on the obtained SAR complex pattern, wherein the InSAR imaging process comprises image registration, phase filtering and phase unfolding based on cluster analysis. The processed InSAR phase unfolded image is processed with an elevation inversion to recover a three dimensional digital elevation map. The invention has the advantages of wide adaptability to mapped terrains, and high imaging effectiveness, therefore, the invention can be used in the mapping of the 3D terrain.

The invention provides a bird voice recognition technology based on novel noise-proof feature extraction by aiming at the problem of bird voice recognition in various kinds of background noise in ecological environment. The bird voice recognition technology comprises the following steps of firstly, obtaining noise power spectrums by a noise estimation algorithm suitable for highly nonstationary environment; secondly, performing the noise reduction on the voice power spectrums by a multi-band spectral subtraction method; thirdly, extracting anti-noise power normalization cepstrum coefficients (APNCC) by combining the voice power spectrums for noise reduction; and finally, performing contrast experiments under the conditions of different environments and signal to noise ratios (SNR) on the voice of 34 species of birds by means of extracted APNCC, power normalization cepstrum coefficient (PNCC) and Mel frequency cepstrum coefficients (MFCC) by a support vector machine (SVM). The experiments show that the extracted APNCC have a better average recognition effect and higher noise robustness and are more suitable for bird voice recognition in the environment with less than 30 dB of SNR.

The invention discloses a structure damage early warning method based on multi-point sensor data and BiLSTM, and the method comprises the steps: arranging a plurality of sensors on a structure, collecting the sensor data in a health state and a damage state of the structure, and building a sample library with a damage value label; normalizing the sensor data in the sample library, and determininga time step length input into the BiLSTM network; building a BiLSTM network model, and training and testing the model; and inputting the multi-point sensor data monitored within a certain period of time into the built BiLSTM network model to obtain a final prediction result of the BiLSTM network model, and judging whether to perform damage early warning or not according to the final prediction result. The method does not depend on a structural finite element model, does not need manual participation in the implementation process, is suitable for automatic online damage early warning of an in-service structure, can give an alarm at the first time when the structure is damaged, and provides certain guidance for bridge maintenance and management decision making, so that safe operation of an engineering structure is guaranteed.

The invention provides a channel estimation method and device. Singular value decomposition is performed on an original observation matrix, according to a result of the singular value decomposition, a reconstructed observation matrix is obtained, sparsity is estimated, the sparsity is the number of nonzero elements in a union of a preset channel impulse response value and an index set omega, wherein omega is generated according to the reconstructed observation matrix and a preset residual vector, and finally, channel estimation is performed using the sparsity. Since the singular value decomposition has the characteristic of relatively strong noise robustness, the method and the device provided by the invention have relatively strong noise robustness, in addition, the method and the device can perform estimation of sparsity without the need to know the sparsity, and thus have relatively high practicability, and the channel estimation method and the device provided by the embodiment have relatively high performance.

The invention discloses a face representation attack detection method based on an LBP-VAE anomaly detection model. The method comprises the steps of constructing a LBP-VAE anomaly detection model, obtaining a training sample, wherein the training sample only needs a real sample, extracting LBP features from the training sample to obtain a sample feature vector, taking the feature vector of the training sample as the input of VAE, and training a VAE network to obtain a complete LBP-VAE anomaly detection model, wherein when a human face representation attack sample is input into the model, the output error of the VAE network is very large and will be detected as anomaly due to the fact that sample feature space distribution is different from that of a real sample, and when the real sample isinput into the model, the output error of the VAE network is small, so that two types of samples can be distinguished. The attack detection method disclosed by the invention has good detection performances for different types of face representation attack samples, is strong in noise robustness, and can adapt to different real scenes.

The invention discloses a high spectral image compression sensing method based on manifold structuring sparse prior and solves a technical problem of low precision existing in a high spectral image compression sensing method in the prior art. The method is characterized in that a few linear observation values of each pixel spectrum are sampled randomly and are taken as compression data, through the manifold structuring sparse prior, sparsity of a high spectral image after sparsification in the spectrum dimension and manifold structure of the high spectral image in the space dimension are etched, through a hidden variable Bayes model, signal reconstruction is carried out, and sparse prior learning and noise estimation are unified to one regularization regression model for optimization solution. The sparse prior acquired through learning can not only fully describe the three-dimensional structure of the high spectral image, but also has relatively strong noise robustness. The sparse prior is utilized to realize high precision reconstruction of the high spectral image. Based on tests, Gauss white noise is added to the compression data to make the signal to noise ratio of the compression data to be 15db, the sampling rate is 0.09, and thereby the 23db peak value signal to noise ratio is acquired.

The invention discloses a time-frequency analysis method based on nonlinear mode decomposition and adaptive optimal kernel; the time-frequency analysis method combines the advantages in nonlinear mode decomposition analysis method and those in adaptive optimal kernel and comprises: decomposing a multi-component nonstable signal into a group of physically significant nonlinear mode components by using nonlinear mode composition algorithm, wherein the requirement on the anti-noise performance of subsequent analysis method is lowered since the algorithm has high noise robustness; then, enabling a kernel function to change adaptively with signal changes by using the time-frequency analysis method of adaptive optimal kernel, so that cross terms are effectively inhibited and time-frequency concentrating capacity is improved. The novel analysis method inherits the advantages in both nonlinear mode composition method and adaptive optimal kernel analysis method, and has excellent performances.

The invention discloses a cell phone source recognizing method under an additive noise environment based on fusion features. The method includes the steps of using the fusion features composed of MFCCfeatures extracted from a Fourier domain, STFTSDF features and CQTSDF features extracted from a CQT domain as device fingerprints so that device differentiation information can be more precisely represented compared with a single feature; training an obtained M classification model to have universality through a multi-scene training mode in the training stage and conducting effective cell phone source recognizing on voice samples in known noise scenes and unknown noise scenes, wherein a training group contains clean voice samples without scene noises and also contains noise-containing voice samples with different scene noise types and noise intensities. The M classification model is established through a deep learning CNN model, and by means of the CNN model, the source recognizing accuracy of the clean voice samples without the scene noises is improved, the cell phone source recognizing effect of the noise-containing voice samples is greatly improved, and noise robustness is high.

The invention provides a power cable partial discharge positioning method. The method comprises the steps that S1 a partial discharge signal of a test cable is acquired, wherein the partial dischargesignal comprises an incident wave signal and a reflected wave signal; S2 the acquired partial discharge signal is decomposed; when the number of times of decomposition is equal to the set number of decomposition layers, the decomposition is stopped; a sub-band signal corresponding to the number of times of decomposition is acquired; S3 the kurtosis of the sub-band signal is calculated, and a firstsub-band signal whose kurtosis is greater than a set kurtosis threshold is acquired; S4 the wavefront position propagation time difference of the incident wave and the reflected wave of each first sub-band signal and the wave velocity corresponding to the center frequency are calculated respectively; S5 according to the propagation time difference and the wave velocity, the partial discharge position of each first sub-band signal is acquired; and S6 the final partial discharge position is acquired according to the partial discharge position of each first sub-band signal. According to the invention, the accuracy of partial discharge positioning can be improved.

The invention belongs to the field of signal processing, and particularly relates to a two-dimensional inversion-free sparse Bayesian learning rapid sparse reconstruction method. The method comprisesthe following steps of: S1: carrying out sparse representation modeling on a two-dimensional sparse reconstruction problem; S2: performing statistical modeling on a vectorized sparse signal x and vectorized noise n; S3: solving the posterior probability of the vectorized sparse signal x, the vectorized inverse variance gamma and the noise inverse variance alpha; and S4: updating the matrix form Zof the auxiliary variables. Compared with the IFSBL method, the two-dimensional inversion-free sparse Bayesian learning rapid sparse reconstruction method has the advantages that the two-dimensional signals are directly processed, the problem that a large matrix is generated due to vectorization of the two-dimensional signals is avoided, the operation efficiency is obviously improved, and the requirement on a calculation memory is obviously reduced; and on the other hand, sparse reconstruction is realized under a statistical signal processing frame, and compared with a non-statistical sparse reconstruction method, the method has the advantages of being easier to obtain a global optimal solution, higher in noise robustness, lower in dependence degree of algorithm performance on parameter initialization and the like, and the engineering practicability is high.

The invention discloses a synthetic aperture radar (SAR) target image geometric size estimation method based on orthogonal matching tracking algorithm OMP. (2) obtaining the atomic matrix and the scattering center amplitude vector of the target; (3) obtaining the reconstructed target image; (4) projecting the reconstructed target image onto the coordinate system to generate the target region; (5)obtaining a coordinate vector; (6) rotating the coordinate vector; (7) obtaining the smallest circumscribed rectangle; (8) The size value of the smallest circumscribed rectangle is used as the estimate value of the target geometric size. The invention is universally applicable to stationary radar observation targets, and at the same time, the invention eliminates the negative influence caused by noise, has good noise robustness in detail, and improves the estimation accuracy of synthetic aperture radar SAR target geometric size.

The invention relates to power load data processing, in particular to a power load data dimension reduction reconstruction processing method based on VMD and OMP. The method comprises the following steps: acquiring power load data; carrying out data decomposition filtering and dimension reduction processing on the acquired power load data through a variational mode decomposition method; and then,after obtaining intrinsic mode components of which the frequencies are from low to high, performing reconstruction optimization processing on the decomposed and filtered data by using an orthogonal matching tracking algorithm. According to the method, a VMD operation method and an OMP operation method are combined for the first time, the dimension reduction reconstruction processing method which is high in operation efficiency and universality is obtained, and the method is suitable for a data set of an existing dimension, not prone to being affected by noise, high in stability and more suitable for power load big data processing derived in the increasingly flourishing development information age.

One example embodiment provides a system, apparatus, and method for using polynomial models in tone calibration for quadrature error correction in I/Q receivers. In one example embodiment, method for calibrating an I/Q receiver is provided and includes receiving a first mismatch parameter indicating a mismatch between I and Q channels of the I/Q receiver; and estimating a second mismatch parameter from the first mismatch parameter using a polynomial model.

The invention discloses an intelligent damage identification method for a jacket type ocean platform. According to the intelligent damage identification method, modal parameters are extracted by adopting a CP algorithm. The adaptability of the algorithm for processing high-damping and high-order complex modes is higher than that of a traditional blind source separation algorithm, such as an ICA algorithm; the CP algorithm adopted by the invention has a direct and effective implementation mode, parameters do not need to be adjusted in the application process, the length of a system vibration response signal does not affect the result accuracy, and complete blind identification can be carried out on modal parameters of various structures; according to the PSO-SVM model provided by the invention, under a small sample condition, the accuracy of a damage identification result is better than that of a traditional intelligent algorithm such as an SVM and a BP neural network, and a model hyper-parameter does not need to be manually adjusted so that a use threshold is reduced; the method provided by the invention has relatively high noise robustness. Experiments show that the method has high identification accuracy under the condition that the noise-to-signal ratio is less than 17.8%.

The invention discloses a face representation attack detection method based on full-size depth map supervision. The method comprises the following steps: constructing a face representation attack detection model based on full-size depth map supervision; inputting a to-be-detected face picture into the trained face representation attack detection model, acquiring a full-size prediction depth map which is identical to the to-be-detected face picture in width and height, wherein each pixel value on the full-size prediction depth map is a depth prediction value of a corresponding pixel point of the to-be-detected face picture; and taking an average depth prediction value of the full-size prediction depth map as a final score, and comparing the final score with a preset discrimination threshold to obtain a detection result. The full-size depth map is adopted as a label for supervised modeling, the shallow-layer features and the deep-layer features of the input face image are fused to obtain the full-size prediction depth map, the detection precision is high, and the invention can adapt to reality detection scenes under different illumination and acquisition equipment conditions.