ML based modulation format identifier using K-NN algorithm

Abstract

This paper uses Machine Learning (ML) to detect PSK (Phase-shift keying) signals using a supervised learning model based on K-Nearest Neighbor (K-NN) algorithm. The incoming signal constellation created has been compared to the training data by the K-NN algorithm to determine the modulation format of the received information. By calculating the Euclidean distance over the whole training set for K occurrences of the nearest data points, four PSK signals: BPSK, QPSK, 8PSK, and 16PSK, have been correctly predicted with a very high accuracy level.