Automatic Modulation Recognition using Convolutional Neural Networks
Muhammad Sabih Ul Qamar, Muhammad Awais Akhter, Rab Nawaz · 2022 19th International Bhurban Conference on Applied Sciences and Technology (IBCAST) · 2022
Automatic modulation classification (AMC), an intermediate step between signal detection and modulation, is an active research area in wireless communication. AMC methodologies utilize the time varying temporal characteristics of complexed in-phase (I) and quadrature (Q) signal to identify the modulation type correctly and automatically. In this work, majority voting based convolution neural network (CNN) is used to recognize modulation scheme. To learn inherent properties more accurately and reduce complexity of the neural network, the signal length is reduced from 128 points to 64. In this regard, a sliding widow of length 64 is used to divide each signal into 5 sub-signals. Instantaneous amplitude (a) and phase (θ) are additionally used as input to the network. Finally, modulation scheme of the signal is decided based on majority voting and is calculated by argument maximum of sub-signal mean. Experiments show that majority voting increases the classification accuracy on publicly available dataset.