Wu Zhilu

Automatic digital modulation recognition using artificial neural networks

This paper presents a modified structure and learning algorithm of artificial neural networks (ANN) for recognizing baseband signal modulation types in the presence of additive white Gaussian noise. The new method employs a layer with less output nodes and an error back propagation learning algorithm with momentum to improve the recognition performance. Simulation results and performance evaluation of the ANN are given and it is shown that the benefits of the developed method are that its structure is simple and it performs well at low signal to noise ratio (SNR) with high overall success rates.

Channelized receiver platform of SDR based on FPGAs

In this paper, a channelized receiver platform of software defined radio is developed which is mainly to realize wideband quadrature digitization, modulation classification and demodulation exploiting A/D converters. FPGAs with the combination of programmable DSPs. The front-end of researched receiver based on polyphase filter quadrature digitization processors, where digitization takes place at intermediate frequencies (IF), adopts a wideband design with a bandwidth of up to 100MHz and a centered carrier frequency of up to 250MHz. The low-complexity high-speed digital IF signal processing is carried out with CORDIC algorithm on FPGAs, then to realize the recognition of several modulation schemes such as 2ASK, 4ASK, 2PSK, 4PSK, 2FSK, 4FSK, 16QAM to implement communication protocols fairly flexibly.In this paper, a channelized receiver platform of software defined radio is developed which is mainly to realize wideband quadrature digitization, modulation classification and demodulation exploiting A/D converters. FPGAs with the combination of programmable DSPs. The front-end of researched receiver based on polyphase filter quadrature digitization processors, where digitization takes place at intermediate frequencies (IF), adopts a wideband design with a bandwidth of up to 100MHz and a centered carrier frequency of up to 250MHz. The low-complexity high-speed digital IF signal processing is carried out with CORDIC algorithm on FPGAs, then to realize the recognition of several modulation schemes such as 2ASK, 4ASK, 2PSK, 4PSK, 2FSK, 4FSK, 16QAM to implement communication protocols fairly flexibly.

A new approach of 2D discrete cosine transform with Mobius inverse formula

The 2D discrete cosine transform (DCT) is an important mathematical tool in digital image processing and many other fields. Our aim is mainly to present an algorithm of 2D DCT, which turns N/spl times/N DCT into N separate 1D DCT and addition operations, and then to implement DCT with an arithmetic Fourier transform (AFT), which has turned out to be an important alternative to the known traditional methods of fast Fourier transform (FFT) in terms of accuracy, complexity and speed. The total number of multiplications and additions is only half of that required for the corresponding traditional method. Simulation results show that this method is feasible and its computational complexity is lower than the row-column algorithm.

Characterization of harmonic distortion and memory effects for RF power amplifiers based on volterra algorithm

The characterization of RF power amplifiers can benefit from the classic Volterra series theory. As a frequency domain form of Volterra series, Volterra algorithm is constituted by Volterra basis and Volterra coefficients. In this paper, the process of extracting Volterra coefficients is proposed and simulated with the harmonic balance simulator. A five order Volterra algorithm has been used to predict time domain waveforms, and a better agreement has been achieved, contrasting with X-parameter model. As an application, two Volterra coefficients concerned with three-order intermodulations is useful for characterizing memory effects influenced by modulation bandwidth.

Nonlinear behavioral modeling and simulation of Rf device with memory effects based on Volterra series

This paper is aiming at making behavioral model of RF device by using Volterra series theory. When we consider the periodic input signal, the Volterra series will presents a discrete form. In this paper, a five-order nonlinear system is applied to derive the Volterra series expansions which can express DC term and harmonic components in frequency domain. The simulation shows that the performance of this behavioral modeling method is good, especially when higher order Volterra series expansion is used. We also conclude that the method given by this paper is valid to both small signal and large signal.

Adaptive neural networks-fuzzy reasoning information fusion system

This paper presents an adaptive neural networks-fuzzy reasoning information fusion system which is employed to decrease the influence of the uncertainty of sensor state on fusion performance under a complex environment. The model consists of a confidence estimator, fusing weight knowledge base and weighted fusion. When utilizing the new model, we can obtain the confidence of each sensor and the reliable fusion data with an adaptive synthetical computation of sensor state, sensor detection precision and track filtering error based on the techniques of neural networks, fuzzy reasoning and knowledge base. The simulation results show that the new fusion model is obviously advantageous compared with the conventional Kalman weighted fusion.

A study on implementing wavelet transform and FFT with FPGA

This paper shows how to implement the wavelet transform and the Fast Fourier Transform (FFT) with a Field Programmed Gate Array (FPGA). First, it introduces the implementation of the wavelet transform with lattice filters and achieves the FFT with the Coordinate Rotational Digital Computation (CORDIC). Then, the emulation data of the Daubechies D4 & D6 wavelet transforms and the FFT with 16 points are given, and their performances are analyzed. The results prove this to be a novel and effective method to implement the wavelet transform and the FFT with FPGA.

A decision-level data fusion based multiuser detection method for DS-UWB systems

In this paper, the decision-level data fusion techniques are extended to the multiuser detection (MUD) field. Then a novel MUD algorithm that is the combine application decision-level fusion criterion (CA-DFC) based MUD algorithm is proposed for DS-UWB communication systems. In CA-DFC based method, the combine detectors can make decisions among the preliminary decisions from sub-optimal detectors by a special rule. Simulation results demonstrate that the performances of CA-DFC based MUD algorithms are superior to those of other sub-optimal MUD algorithms, and even close to that of OMD. Moreover, the proposed algorithm have lower computational complexity than OMD, which reveals their efficiency. Compared with CA-DFC based algorithm achieves a little improvement in its performance, at the cost of the increment in its computational complexity. Thus, they can be applied to different practical situations.

A low complexity iterative multiuser detector for LDPC-coded TDRSS return link MA service

In coded multiuser tracking and data relay satellite system (TDRSS) return link, multiple access interference and inter-symbol interference are major impediment to reliable communication. While the optimal soft-input soft-output multiuser detection scheme results in a prohibitive computational complexity. In this paper, a simplified hard-input soft-output (HISO) scheme for low-density parity-check (LDPC) coded multiuser TDRS system is proposed. The hard decisions of multiuser detection are used as the initial values of the decoding method. After each iteration, the hard output of log-likelihood ratio (LLR) belief propagation (BP) decoding algorithm is fed back as the input of decoding for the next iteration. Furthermore, interleaving is not a necessary part in the proposed method. The computational complexity is much lower than the optimal algorithm. However, simulation results show that the performance of the proposed scheme approaches that of the single-user system for moderate to high signal-to-noise ratios (SNRs).

Cross-Layer Power Control of Ad Hoc Network Based on Ant Colony Optimization

This paper studies the power control mechanism of MAC layer and network layer of ad hoc network and proposes a cross-layer power control model which absorbs ant colony optimization in it. After getting the minimum sending power between each two nodes, the virtual central node can find network topology by Ant Colony Optimization (ACO) in network layer, then each pair of communication nodes adaptively adjust the current optimal sending power according to the exchange of information between nodes in link layer to realize the cross-layer power control. The simulation results show that in ad hoc network, the power control mechanism can quickly find the optimal path and effective reduce the energy and power consumption of nodes.