AM-DSB MODULATION DETECTION AMONG SIGNALS MODULATED WITH 26 DIFFERENT MODULATION TECHNIQUES WITH MobileNet ARCHITECTURE


Aytekin A., Mençik V., Budak C.

International Informatics Congress (IIC2022) , Batman, Turkey, 17 - 19 February 2022, pp.123-129

  • Publication Type: Conference Paper / Full Text
  • City: Batman
  • Country: Turkey
  • Page Numbers: pp.123-129
  • Abdullah Gül University Affiliated: Yes

Abstract

 The idea of optimizing communication systems using deep learning techniques is promising for the future. This new learning method is expected to reduce the need for high power consuming electronic circuits for solving complex mathematical expressions in communication systems. The use of deep learning techniques for modulation sensor designs, which is one of the points in need of improvement in communication systems, will contribute to the literature. In this study, it is aimed to design a CNN-based modulation detector that detects AM-DSB modulated signals among 26 different modulation techniques, each of which has equal SNR values, by using the HisarMod2019 data set. MobilNet architecture was used for experimental studies. With this architecture, 100% F1-score value was obtained. This result shows that the used architecture has the ability to recognize AM-DSB samples in the correlation between classification ability.