Modulation Recognition of Radar Signal Based on an Improved CNN Model

Cai, Jingjing; Li, Chao; Zhang, Huanyin

doi:10.1109/iccsnt47585.2019.8962418

Cited by 9 publications

(7 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Figures 5 shows the confusion matrix of the CLSTM network at different SNR. From the figure, it can be seen that the network is poor in recognizing all modulated signals at low SNR, because the interference of noise blurred the features of IQ signals [15] ; At medium to high SNR, the recognition accuracy is high for all modulations except for modulations QAM16 and QAM64, AM-DSB and WBFM which are easily confused [16] . Therefore, the next step can be divided into two directions: First is to do data enhancement, noise reduction and other preprocessing on the original data to improve the feature extraction ability of the network at low SNR; Second is to enhance the intra-class recognition ability of the network to improve the discrimination effect of the network on the confused signals [17] .…”

Section: Recognition Effect Of Clstm Network With Different Snrmentioning

confidence: 99%

A CLSTM network algorithm for automatic modulation recognition

Wang

Fang

Fan

et al. 2023

International Conference on Signal Processing, Computer Networks, and Communications (SPCNC 2022)

View full text Add to dashboard Cite

Aiming at the problems of low recognition accuracy and inexplicable features extracted by neural networks in signal modulation recognition algorithms based on deep learning, this paper proposes a CLSTM recognition algorithm based on IQ signal features. The algorithm first uses a 2×1 convolution kernel to extract the phase feature of the signal, then uses a multiple convolution layer to extract the temporal feature of the signal, and then outputs the result of the convolution layer to the LSTM module to further extract the temporal feature, finally, the recognition result is output by the fully connected layer and the recognition effect is improved by optimizing the network structure. The experimental results show that on the RML2016.10a public dataset, when the SNR is higher than 0 dB, the recognition accuracy of the algorithm reaches 89.21%, which is 15.02% and 9.85% higher than the classical CNN2 and CLDNN network algorithms.

show abstract

Section: Recognition Effect Of Clstm Network With Different Snrmentioning

confidence: 99%

A CLSTM network algorithm for automatic modulation recognition

Wang

Fang

Fan

et al. 2023

International Conference on Signal Processing, Computer Networks, and Communications (SPCNC 2022)

View full text Add to dashboard Cite

show abstract

“…AirNN closely follows the bound of the software-based RQM-CNN, with a drop in accuracy of only 2%, and an overall drop w.r.t. Classic-CNN of 3.2% for the SNR range of [6,30] dB.…”

Section: Airnn Performancementioning

confidence: 99%

“…(3) As a systems contribution, we implement a softwareframework to control the RIS network called AirNNOS that synchronizes and aligns start times of the relay transmitters and the receiver, as well as reconfigures the RIS on demand to change their reflection coefficients. (4) Given the measured error of the over-the-air convolution, we show through simulations that the experimentally derived analog convolution is accurate enough to run inference on trained neural networks, with an average deviation in testing accuracy of 3.2% for a range of medium-tohigh SNR of [6,30] dB compared to classical, GPU-based inference.…”

Section: Introductionmentioning

confidence: 96%

“…The wireless research community has applied machine learning (ML) methods for physical layer related problems of protocol classification [5], adversarial activity detection, modulation classification [6] and RF fingerprinting [7], among others. In particular, the ML solutions proposed in these works are based on a special class of architectures called as convolutional neural networks (CNNs).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

AirNN: Neural Networks with Over-the-Air Convolution via Reconfigurable Intelligent Surfaces

Sánchez¹,

Muns²,

Bocanegra³

et al. 2022

Preprint

View full text Add to dashboard Cite

Over-the-air analog computation allows offloading computation to the wireless environment through carefully constructed transmitted signals. In this paper, we design and implement the first-of-its-kind over-the-air convolution and demonstrate it for inference tasks in a convolutional neural network (CNN). We engineer the ambient wireless propagation environment through reconfigurable intelligent surfaces (RIS) to design such an architecture, which we call 'AirNN'. AirNN leverages the physics of wave reflection to represent a digital convolution, an essential part of a CNN architecture, in the analog domain. In contrast to classical communication, where the receiver must react to the channel-induced transformation, generally represented as finite impulse response (FIR) filter, AirNN proactively creates the signal reflections to emulate specific FIR filters through RIS. AirNN involves two steps: first, the weights of the neurons in the CNN are drawn from a finite set of channel impulse responses (CIR) that correspond to realizable FIR filters. Second, each CIR is engineered through RIS, and reflected signals combine at the receiver to determine the output of the convolution. This paper presents a proof-of-concept of AirNN by experimentally demonstrating over-the-air convolutions. We then validate the entire resulting CNN model accuracy via simulations for an example task of modulation classification.

show abstract

“…Through the radar signal preprocessing and feature extraction of the convolutional neural network, the network can identify random overlapping radar signals under low SNR. Cai et al [ 9 ] proposed a radar signal modulation and recognition algorithm based on an improved CNN model. In this model, a dense connection block layer and a global pooling layer were added to identify 8 radar signals.…”

Section: Introductionmentioning

confidence: 99%

A Radar Signal Recognition Approach via IIF-Net Deep Learning Models

Zhang

et al. 2020

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

In the increasingly complex electromagnetic environment of modern battlefields, how to quickly and accurately identify radar signals is a hotspot in the field of electronic countermeasures. In this paper, USRP N210, USRP-LW N210, and other general software radio peripherals are used to simulate the transmitting and receiving process of radar signals, and a total of 8 radar signals, namely, Barker, Frank, chaotic, P1, P2, P3, P4, and OFDM, are produced. The signal obtains time-frequency images (TFIs) through the Choi–Williams distribution function (CWD). According to the characteristics of the radar signal TFI, a global feature balance extraction module (GFBE) is designed. Then, a new IIF-Net convolutional neural network with fewer network parameters and less computation cost has been proposed. The signal-to-noise ratio (SNR) range is −10 to 6 dB in the experiments. The experiments show that when the SNR is higher than −2 dB, the signal recognition rate of IIF-Net is as high as 99.74%, and the signal recognition accuracy is still 92.36% when the SNR is −10 dB. Compared with other methods, IIF-Net has higher recognition rate and better robustness under low SNR.

show abstract

Modulation Recognition of Radar Signal Based on an Improved CNN Model

Cited by 9 publications

References 9 publications

A CLSTM network algorithm for automatic modulation recognition

A CLSTM network algorithm for automatic modulation recognition

AirNN: Neural Networks with Over-the-Air Convolution via Reconfigurable Intelligent Surfaces

A Radar Signal Recognition Approach via IIF-Net Deep Learning Models

Contact Info

Product

Resources

About