Design of the Image-Based Deep Learning Model Using a Pre-Training Deep Learning Network 


Vol. 47,  No. 4, pp. 615-624, Apr.  2022
10.7840/kics.2022.47.4.615


PDF
  Abstract

In this paper, an image-based deep learning model is designed for Automatic Modulation Classification (AMC) in cognitive radio. The proposed design method consists of a signal-based deep learning model and an image-based deep learning model, and a Convolutional Neural Network (CNN) is used for the deep learning network type. The signal-based deep learning model is trained in units of a frame which is composed of 1024 signal samples. Before being used for training, the frame is normalized through the Root Mean Square (RMS) method, and the frame is dividied into the real part and the imaginary part. The proposed signal-based deep learning model is analyzed according to the filter size of the convolution layer and optimized by specifying 1×8 filter size. The image-based deep learning model is trained through images that is from the extracted feature data using the pretrained signal-based deep learning network, and predicted each modulation type. The feature size is 24×1, which is extracted through the Fully-Connected layer of the signal-based deep learning network, and the features are converted into Red, Green, and Blue (RGB) images according to the -30 - +30 scale range. The prediction performance of the proposed model shows 2.13%, 4.05% and 9.47% higher accuracy at Signal-to-Noise Ratio (SNR) 10 ㏈, and 2.12%, 3.4% and 4.26% higher accuracy at SNR 0 ㏈ than the conventional models ECNN, SCGNet and MCNet, respectively.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Cite this article

[IEEE Style]

S. Kim, C. Moon, K. Kwon, D. Kim, "Design of the Image-Based Deep Learning Model Using a Pre-Training Deep Learning Network," The Journal of Korean Institute of Communications and Information Sciences, vol. 47, no. 4, pp. 615-624, 2022. DOI: 10.7840/kics.2022.47.4.615.

[ACM Style]

Seung-Hwan Kim, Chang-Bae Moon, Ki-Hyeob Kwon, and Dong-Seong Kim. 2022. Design of the Image-Based Deep Learning Model Using a Pre-Training Deep Learning Network. The Journal of Korean Institute of Communications and Information Sciences, 47, 4, (2022), 615-624. DOI: 10.7840/kics.2022.47.4.615.

[KICS Style]

Seung-Hwan Kim, Chang-Bae Moon, Ki-Hyeob Kwon, Dong-Seong Kim, "Design of the Image-Based Deep Learning Model Using a Pre-Training Deep Learning Network," The Journal of Korean Institute of Communications and Information Sciences, vol. 47, no. 4, pp. 615-624, 4. 2022. (https://doi.org/10.7840/kics.2022.47.4.615)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880