Aiming at the problem that square integral bispectrum cannot accurately represent communication transmitter individuals with the same type under small sample, a communication transmitter individual deep square integra...
详细信息
Aiming at the problem that square integral bispectrum cannot accurately represent communication transmitter individuals with the same type under small sample, a communication transmitter individual deep square integral bispectra feature extraction method based on stacked autoencoder network is proposed. The square integral bispectrum (SIB) features of real communication transmitters signals were extracted firstly from the instantaneous frequency. Then the deep SIB features were represented through the layer-wise semi-supervised learning by a stacked autoencoder network. Experiments on real FM communication transmitter signals showed that the identification accuracy is around 80% and the results of multiple experimental variables setting indicated the robustness of the algorithm proposed.
作者：
Zheng, YangfengSUN YAT SEN UNIV
Sch Data & Comp Sci Guangzhou Higher Educ Mega Ctr Univ Town 132 East Waihuan Rd Guangzhou 510006 Peoples R China
As the size of training dataset of face recognition models becomes larger and larger, we are interested in a method called Average-Half-Face (AHF), which could halve the size of training samples. The AHF method divide...
详细信息
As the size of training dataset of face recognition models becomes larger and larger, we are interested in a method called Average-Half-Face (AHF), which could halve the size of training samples. The AHF method divides a full face into two halves and then averages them together (reversing the columns of one of the halves). We preprocess the dataset with the method of AHF, and train them on two different models, Eigenfaces and Convolutional Neural Network (CNN). We compare the prediction results with those models trained on the original dataset. Previous researches showed that AHF is superior to Full-Face (FF), while our experiment results further showed that in some cases AHF also boosts CNN. The application of AHF can bring both saving in storage and reduction on training cost time.
Frequency diverse array (FDA) which employs a small frequency increment between adjacent elements can provide a range-angle-dependent beampattern with new application potentials. In this paper, the proposed coherent p...
详细信息
Frequency diverse array (FDA) which employs a small frequency increment between adjacent elements can provide a range-angle-dependent beampattern with new application potentials. In this paper, the proposed coherent pulsed-LFM FDA scheme enables full spatial coverage with a single transmit waveform, wherein the transmit beamforming can be performed at the receiver with flexible signalprocessing means. In the coherent pulsedLFM FDA radar configuration, the range sidelobe level (SLL) can be reduced significantly. Analyses of the multi-dimensional ambiguity function in terms of the low sidelobe characteristics, spatial coverage capability, and resolution properties in range, angle and Doppler domains are given by ambiguity function profiles. Comparisons with conventional phased array and MIMO (Multiple-Input Multiple-Output) radar are presented in the simulation. Results demonstrate the superiority of the proposed approach in angular coverage and sidelobe level with simplicity in engineering.
In the 5th generation mobile communication (5G) system, it is required to provide three different types of services, including enhanced Mobile Broadband (eMBB), massive machine-type communication (mMTC), and ultra-rel...
详细信息
In the 5th generation mobile communication (5G) system, it is required to provide three different types of services, including enhanced Mobile Broadband (eMBB), massive machine-type communication (mMTC), and ultra-reliable and low-latency communication (URLLC).To meet all those demand in 5G, the waveform is crucial. Although the CP-OFDM waveform has been wildly used in 4G, it is hardly to satisfy the requirements of 5G.A lot of candidate waveforms are discussed for 5G system, including UFMC, FBMC, F-OFDM etc. In this paper, the UFMC waveform is discussed, the different filters and different frequency width of the subband in UFMC system are studied and appropriate parameters for the system are recommended.
Due to the good adaptability of wavelet theory in signalprocessing, its current application fields are very extensive. This paper first summarizes the application fields of current wavelet theory and analyzes the cha...
详细信息
Due to the good adaptability of wavelet theory in signalprocessing, its current application fields are very extensive. This paper first summarizes the application fields of current wavelet theory and analyzes the characteristics of wavelet analysis theory. Then, several methods for fault diagnosis of traditional inverter circuits are listed. The limitations of their fault diagnosis in new multi-level inverter circuits are analyzed. The wavelet analysis is applied to the fault diagnosis of multi-level inverter circuits. After analyzing the fault mode of the multi-level inverter circuit, the wavelet packet is used to extract the fault features, and then the support vector machine theory is combined to diagnose the fault. Finally, the method described in this paper is verified by Simulink simulation software, and the accuracy of the method is more than 90%. The method is feasible and effective.
In the process of signal transmission of offset quadrature amplitude modulation/orthogonal frequency division multiplexing (OQAM/OFDM) system, high power amplifier (HPA) causes signal distortion due to the limited amp...
详细信息
In the process of signal transmission of offset quadrature amplitude modulation/orthogonal frequency division multiplexing (OQAM/OFDM) system, high power amplifier (HPA) causes signal distortion due to the limited amplitude of high power signal, which lead to a decrease of capacity of the system. To calculate the system capacity under the effect of HPA, according to the Bussagang theorem, the equivalent signal-to-noise ratio (ESNR) of the signal under HPA is derived. The validity of the derivation is verified by the bit error rate (BER) performance of the system. The capacity mean and capacity variance with the effect of HPA of the system are obtained according to the derivation in the Rayleigh fading channel. This method provides reference for performance analysis and peak to average power ratio (PAPR) reduction of OQAM/OFDM system.
Transfer learning aims to transfer knowledge from the labelled source domain data to build a good classifier for the target domain data which has few labels or none. Existing feature-based transfer learning methods se...
详细信息
Transfer learning aims to transfer knowledge from the labelled source domain data to build a good classifier for the target domain data which has few labels or none. Existing feature-based transfer learning methods seek to transform the data to a new feature space under which the distribution discrepancy is reduced. However, data in different classes may not be easy to be separated in the new feature representation. Therefore, a modified transfer learning algorithm with Joint Distribution Adaptation (JDA) and Maximum Margin Criterion (MMC) is put forward in this paper, which we call MMC-JDA for short. MMC-JDA is aimed at minimizing the distribution discrepancy between two domains and maximizing the separability of each class at the same time. Comparative experiments on 16 cross-domain public image datasets show that MMC-JDA is effective and performs better than several common transfer learning methods.
The influence of residual carrier frequency offset on demodulation for multi-tone parallel signals is theoretically deduced and simulated, and two compensation algorithms for residual carrier frequency offset are prop...
详细信息
The influence of residual carrier frequency offset on demodulation for multi-tone parallel signals is theoretically deduced and simulated, and two compensation algorithms for residual carrier frequency offset are proposed. Taking 16 tones for example, the simulation results in AWGN channel show that the influence of the residual carrier frequency offset is in agreement with the theoretical derivation, and the loss of error code performance is less than 0.1db when the frequency offset is less than 0.4% of the sub-carrier symbol rate. The thresholds of the SNR lossing less than 0.1db are obtained: Eb/N0=5db for short burst signals and Eb/N0=4db for continuous signals. The theoretical analysises and simulation results show that the proposed algorithms are applicable to multi-tone parallel signals.
The problem of stationary target location by multiple passive radar sensors that using unknown and non-cooperative opportunity illuminator is considered. Traditional two-step approach which estimating the time differe...
详细信息
The problem of stationary target location by multiple passive radar sensors that using unknown and non-cooperative opportunity illuminator is considered. Traditional two-step approach which estimating the time difference of arrival (TDOA) and angle of arrival (AOA) firstly and locating using those parameters secondly. We explore the direct location with multiple passive radar sensors without estimating the intermediate parameters. As the reference path from the transmitter to receivers may be blocked in practice, we discuss the both cases that location without and with reference path. Two maximum likelihood algorithms of direct location are proposed for multiple passive radar sensors without and with reference. Monte Carlo simulations indicate that the direct location algorithm is superior to two-step approach with TDOA and AOA at low SNR for multiple passive radar sensors.
In the view of the target dense region, the feartures of different target's time difference are not obvious, which leads to the problem of JTIDS multi-user sorting useless based on time difference. A nonlinear tra...
详细信息
In the view of the target dense region, the feartures of different target's time difference are not obvious, which leads to the problem of JTIDS multi-user sorting useless based on time difference. A nonlinear transformation sorting algorithm of high-order cumulant based on doppler frequency shift is proposed: First of all,make the nonlinear transformation based on the second order cumulant,the sixth order cumulant, the normalized skewness and the normalized kurtosis for single station signals respectively, achiece a sharp rise occurs in the degree of the frequency shift;Then,build the eigenvector according to the transformation;Finally,use spectral clustering algorithm to cluster the datasets. Simulation experiments were carried out,based on four targets which could not be sorted by the time difference. The result showed that compared with the traditional method,the success rate of our method was significantly improved, and the effectiveness of the algorithm was verified.
暂无评论