Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Chen Wu has a total of 62 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A Computational System-Level Evaluation of a Multi-Octave Omnidirectional Direction-Finding System Using Full-Wave Electromagnetic Modeling of a Biconical Array and High-Fidelity Digital Receiver Simulation

    by Chen Wu, Nandha Dileep, Matthew Ruddick

    A deep neural network-based 2D multi-octave omnidirectional direction-finding (DNN-based MOOD-DF) framework was previously developed using seven isotropic antenna elements arranged in a randomly distributed 3D array together with a simplified digital receiver model. In this paper, a system-level evaluation of the corresponding DF system is performed using a full-wave electromagnetic 3D DF array and high-fidelity receiver model. The isotropic elements and simplified receiver models are replaced with 2–18 GHz biconical antenna elements and high-fidelity digital receiver models that account for array feed structure effects, mutual coupling, receiver noise, cable length mismatches, and antenna receiver impedance mismatches within the RF receiving chain. Because full-array multi-octave electromagnetic simulations are computationally intensive, three representative frequency blocks (2–4 GHz, 10–12 GHz, and 16–18 GHz), spanning the lower, middle, and upper portions of the 2–18 GHz frequency range, are selected for DF system performance evaluation. The results demonstrate that the proposed DF system remains effective when trained and tested using the high-fidelity antenna array and receiver models. Across the investigated frequency bands, the system achieves sub-degree azimuth estimation accuracy and elevation RMSEs below approximately 1.8° at SNRs near 0 dB while maintaining microsecond-scale inference latency suitable for real-time applications. Although a performance gap relative to the Cramér–Rao lower bound remains, benchmark comparisons with the spectral MUSIC algorithm at selected frequencies indicate that the proposed approach can successfully operate on the complex and highly spatially aliased 3D manifold of the multi-octave biconical array, whereas conventional narrowband DF techniques may encounter substantial challenges under the same conditions. The presented end-to-end modeling and validation framework of the DF system provides a bridge between previous idealized studies and future hardware prototyping and experimental validation.

  • Open Access

    A New Design Approach to Multi-Octave 2D Omnidirectional Direction-Finding Array With Multi-Band Deep Neural Network Models

    by Chen Wu, Qi Er Teng

    This paper focuses on presenting a new design methodology for a multi-octave omnidirectional direction-finding (MOOD-DF) array tailored for ultra-wideband digital receivers operating across a broad spectrum, e.g., 2 to 18 GHz. The system features a compact, randomly positioned 7-element 3D array that provides full 360° azimuth and defined elevation coverage. Direction-of-arrival (DOA) estimation across the 3.17-octave span is performed using 8 sub-band deep neural network (DNN) models, each selected from a pool of uniform multilayer perceptrons optimized for its frequency partition. Each DNN receives inputs formed by concatenating mapped frequency with the real and imaginary components of array steering vectors, enabling robust frequency representation and distinctive feature encoding. The random array geometry, sub-band modeling, and mapped frequency input eliminate spatial (frequency) aliasing while supporting efficient training, validation, and inference. For the first time in DNN-based RF DOA estimation, each DNN model outputs the three components of a Cartesian unit direction vector, which improves training stability and accelerates convergence compared to angle-based outputs. Training samples were generated at 10 dB signal-to-noise ratio (SNR) to enhance generalization under realistic conditions. Using batch-data inference on an NVIDIA RTX 6000 Ada GPU, the proposed method achieves low single-digit microsecond average estimation time. Simulation results show overall root-mean-square errors below 2° in azimuth and 2.5° in elevation at approximately 0 dB SNR, confirming the system’s high accuracy, speed, and robustness. These results support the integration of the proposed MOOD-DF array into advanced digital receivers for near real-time electronic support measures.


  • Open Access

    Deep Neural Network Model of Ultrafast 2D Direction‐of‐Arrival Estimation Using Planar Arrays for Multi‐Octave‐Band Digital Receiver Applications

    by Chen Wu, Qi Er Teng, Raffi Fox

    This study presents a deep neural network (DNN) model for multi‐octave‐band direction‐finding (MOB‐DF) estimation using a broadband DF‐array and multi‐layer perceptron (MLP). The model leverages randomly placed array elements to generate unique array steering vectors (ASVs) for directions within a cone‐shaped field‐of‐view. By directly linking ASVs and signal frequency to direction via an MLP, it eliminates reliance on the signal covariance matrix, a common component in many 2D neural network‐based DF methods. The DNN‐based MOB‐DF model is structured into sub‐bands, each utilising a trained 16 × 1024 MLP. Simulations with 3‐, 4‐, and 5‐element DF models, trained and validated on datasets with signal‐to‐noise ratios (SNRs) of 10, 20, and 100 dB respectively, reveal several key findings: (1) MLPs trained at 10 dB SNR can achieve better estimation performance across varying SNR levels, where estimation performance is defined as the probability of direction estimation error ≤ 1°. (2) Increasing array elements expands MOB coverage. (3) The 5‐element model attains probabilities of 50% and 90% for ≤ 1° estimation errors at approximately −20 and −10 dB SNR respectively within 2–20 GHz. (4) Average prediction time per direction is on the microsecond scale. (5) The model shows resilience to frequency estimation uncertainties.


  • Open Access

    High sensitivity multi‐channel digital receiver for wideband very weak signal direction‐finding classified by machine learning

    by Chen Wu, Michael Low

    For weak signal detection with direction‐finding (DF), this article presents a new receiver design approach that combines our accumulatively increasing receiver sensitivity (AIRS) signal detection algorithm with the compressive‐sensing (CS)‐based DF‐array/algorithm. The former uses the concept of timeslot (TS)‐based signal threshold detection, whereas the latter employs a frequency‐independent array with randomly located elements, whose bandwidth (BW) largely determines the DF‐array BW. To estimate the direction of a signal, the AIRS algorithm generates the array steering vectors in each TS when the amplitude of any frequency bins exceeds the predetermined threshold of the TS. The aim of this paper is to demonstrate the ability of the new receiver to detect low probability of intercept radar signals with high DF accuracy, fine frequency resolution, and good time‐of‐arrival measurement resolution. To discriminate accurate emitter directions from many false estimations created by the DF‐array in very low signal‐to‐noise ratio environments, K‐means clustering was also applied. In a scenario, the frequency modulated signals from several 165‐mW X‐band radars were in the field of view of a 6‐element DF‐array. Simulation results show that the receiver can accurately estimate all the emitters' directions with root mean squared error of less than 1°, when the separation between the DF‐array and radars is about 100 km.


  • Open Access

    FFT-Based Simultaneous Calculations of Very Long Signal Multi-Resolution Spectra for Ultra-Wideband Digital Radio Frequency Receiver and Other Digital Sensor Applications

    by Chen Wu, Michael Low

    The discrete Fourier transform (DFT) is the most commonly used signal processing method in modern digital sensor design for signal study and analysis. It is often implemented in hardware, such as a field programmable gate array (FPGA), using the fast Fourier transform (FFT) algorithm. The frequency resolution (i.e., frequency bin size) is determined by the number of time samples used in the DFT, when the digital sensor’s bandwidth is fixed. One can vary the sensitivity of a radio frequency receiver by changing the number of time samples used in the DFT. As the number of samples increases, the frequency bin width decreases, and the digital receiver sensitivity increases. In some applications, it is useful to compute an ensemble of FFT lengths; e.g., 2P−j for j=0, 1, 2, …, J, where j is defined as the spectrum level with frequency resolution  2j·Δf. Here Δf is the frequency resolution at j=0. However, calculating all of these spectra one by one using the conventional FFT method would be prohibitively time-consuming, even on a modern FPGA. This is especially true for large values of P; e.g., P≥20. The goal of this communication is to introduce a new method that can produce multi-resolution spectrum lines corresponding to sample lengths  2P−j for all J+1 levels, concurrently, while one long 2P-length FFT is being calculated. That is, the lower resolution spectra are generated naturally as by-products during the computation of the 2P-length FFT, so there is no need to perform additional calculations in order to obtain them.


  • Open Access

    Very Long-Length FFT Using Multi-Resolution Piecewise-Constant Windows for Hardware-Accelerated Time–Frequency Distribution Calculations in an Ultra-Wideband Digital Receiver

    by Chen Wu, Janaka Elangage

    The hardware-accelerated time–frequency distribution calculation is one of the commonly used methods to analyze and present the information from intercepted radio frequency signals in modern ultra-wideband digital receiver (DRX) designs. In this paper, we introduce the piecewise constant window blocking FFT (PCW-BFFT) method. The purpose of this work is to show the generation of spectrograms (formed by a number of spectrum lines) using a very large number of samples (N) in an FFT frame for each spectrum line calculation. In the PCW-BFFT, the N samples are grouped into K consecutive time slots, and each slot has M number of samples. As soon as the M samples in the current time slot are available from a high-speed analog-to-digital convertor (ADC), the frequency information will be obtained using K M-point FFTs. Since each time the FFT frame hops one time slot for the next spectrum line calculation, the frequency information obtained from a time slot will be reused in many spectrum line calculations, as long as these spectrum lines share those samples in the time slot. Although the use of the time domain PCW introduces spikes in the frequency spectrum of the window, the levels of those spikes are still much lower than the first side lobe level of a rectangular window. Using a Gaussian window as an example, the highest spike level can be lower than the main lobe level by at least 38 dB. The PCW-BFFT method allows a DRX to produce multiple spectrograms concurrently with different analysis window widths when the time domain samples become available continuously from the ADC. This paper presents the detailed derivation process of the PCW-BFFT method and demonstrates the use of the method with simulation results. The hardware implementation process will be reported in another paper. The computer simulation results show that long signals with slowly changing frequencies over time can be depicted on the spectrograms with wide analysis windows, and short pulses and signals with rapidly changing instantaneous frequencies can be captured in the narrow analysis window spectrograms.


  • Open Access

    Accumulatively Increasing Sensitivity of Ultrawide Instantaneous Bandwidth Digital Receiver with Fine Time and Frequency Resolution for Weak Signal Detection

    by Chen Wu, Taiwen Tang, Janaka Elangage, Denesh Krishnasamy

    It is always an interesting research topic for digital receiver (DRX) designers to develop a DRX with (1) ultrawide instantaneous bandwidth (IBW), (2) high sensitivity, (3) fine time-of-arrival-measurement resolution (TMR), and (4) fine frequency-measurement resolution (FMR) for weak signal detection. This is because designers always want their receivers to have the widest possible IBW to detect far away and/or weak signals. As the analog-to-digital converter (ADC) rate increasing, the modern DRX IBW increases continuously. To improve the signal detection based on blocking FFT (BFFT) method, this paper introduces the new concept of accumulatively increasing receiver sensitivity (AIRS) for DRX design. In AIRS, a very large number of frequency-bins can be used for a given IBW in the time-to-frequency transform (TTFT), and the DRX sensitivity is cumulatively increased, when more samples are available from high-speed ADC. Unlike traditional FFT-based TTFT, the AIRS can have both fine TMR and fine FMR simultaneously. It also inherits all the merits of the BFFT, which can be implemented in an embedded system. This study shows that AIRS-based DRX is more efficient than normal FFT-based DRX in terms of using time-domain samples. For example, with a probability of false alarm rate of 10−7, for N=220 frequency-bins with TMR = 50 nSec, FMR = 2.4414 KHz, IBW > 1 GHz and ADC rate at 2.56 GHz, AIRS-based DRX detects narrow-band signals at about −42 dB of input signal-to-noise ratio (Input-SNR), and just uses a little less than N/2 real-samples. However, FFT-based DRX have to use all N samples. Simulation results also show that AIRS-based DRX can detect frequency-modulated continuous wave signals with ±0.1, ±1, ±10 and ±100 MHz bandwidths at about −39.4, −35.1, −30.2, and −25.5 dB of Input-SNR using about 264.6 K, 104.7 K, 40.2 K and 18.3 K real-samples, respectively, in 220 frequency-bins for TTFT.


  • Open Access

    Hardware Implementation and RF High-Fidelity Modeling and Simulation of Compressive Sensing Based 2D Angle-of-Arrival Measurement System for 2–18 GHz Radar Electronic Support Measures

    by Chen Wu, Denesh Krishnasamy, Janaka Elangage

    This article presents the hardware implementation and a behavioral model-based RF system modeling and simulation (M&S) study of compressive sensing (CS) based 2D angle-of-arrival (AoA) measurement system for 2–18 GHz radar electronic support measures (RESM). A 6-channel ultra-wideband RF digital receiver was first developed using a PXIe-based multi-channel digital receiver paired with a 6-element random-spaced 2D cavity-backed-spiral-antenna array. Then the system was tested in an open lab environment. The measurement results showed that the system can measure AoA of impinging signals from 2–18 (GHz) with overall RMSE of estimation at 3.60, 2.74, 1.16, 0.67 and 0.56 (deg) in L, S, C, X and Ku bands, respectively. After that, using the RF high-fidelity M&S (RF HF-M&S) approach, a 6-channel AoA measurement system behavioral model was also developed and studied using a radar electronic warfare (REW) engagement scenario. The simulation result showed that the airborne AoA measurement system could successfully measure an S-band ground-based target acquisition radar signal in the dynamic REW environment. Using the RF HF-M&S model, the applicability of the system in other frequencies within 2–18 (GHz) was also studied. The simulation results demonstrated that the airborne AoA measurement system can be used for 2–18 GHz RESM applications.


  • Open Access

    A Signal Processing Algorithm of Two-Phase Staggered PRI and Slow Time Signal Integration for MTI Triangular FMCW Multi-Target Tracking Radars

    by Taiwen Tang, Chen Wu, Janaka Elangage

    In this paper, a novel signal processing algorithm for mitigating the radar blind speed problem of moving target indication (MTI) for frequency modulated continuous wave (FMCW) multi-target tracking radars is proposed. A two-phase staggered pulse repetition interval (PRI) solution is introduced to the FMCW radar system. It is implemented as a time-varying MTI filter using twice the hardware resources as compared to a uniform PRI MTI filter. The two-phase staggered PRI FMCW waveform is still periodic with a little more than twice the period of the uniform PRI radar. We also propose a slow time signal integration scheme for the radar detector using the post-fast Fourier transformation Doppler tracking loop. This scheme introduces 4.77 dB of extra signal processing gain to the signal before the radar detector compared with the original uniform PRI FMCW radar. The validation of the algorithm is done on the field programmable logic array in the loop test bed, which accurately models and emulates the target movement, line of sight propagation and radar signal processing. A simulation run of tracking 16 s of the target movement near or at the radar blind speed shows that the total degradation from the raw post-fast Fourier transformation received signal to noise ratio is about 2 dB. With a 20 dB post-processing signal to noise ratio of the proposed algorithm for the moving target at around a 20 km range and with about a −3.5 dB m2 radar cross section at a 1.5 GHz carrier frequency, the tracking errors of the two-dimensional angles with a 4×4 digital phased array are less than 0.2 degree. The range tracking error is about 28 m.