BLIND MODULATION CLASSIFIER FOR NON COOPERATIVESIGNAL DEMODULATION

Abstract

This is the final report of the project "Blind Modulation Classification for Non-Cooperative Demodulation", conducted by the students of the Department of Computer System Engineering at College Of Electrical & Mechanical Engineering (NUST). The objective of the project was to design and implement Automatic Modulation Recognition system with least dependency on the preprocessing. A feature-based method is used, introducing new intuitive features for Real-Time Classification of Digitally modulated signals without any prior knowledge of signal parameters. The incoming signal's basic modulation type is detected i.e. FSK, PSK, ASK & QAM and then its order is identified. This hierarchical classification can be considered a step towards a general modulation classifier in AWGN channel. Linear Approximations are introduced in Instantaneous Amplitude and Non-Linear Component of Instantaneous Phase which results in improved performance of the system at lower SNR values. Simulations show that with the new feature set classification success rate is 99.9% at very low SNR i.e. 5dB. Estimation algorithms are introduced for signal parameters like Symbol Rate and Carrier Frequency, a new wavelet transform based approach was devised to deal with all basic modulation schemes for the purpose of Symbol Rate Estimation ,the algorithm showed promising result even at low SNR values i.e OdB. Carrier Frequency Estimation was carried out using Daniell filtered smoothed periodogram approach, which is one of the latest techniques in frequency spectrum analysis. The system was designed and Implemented in two phases, in the first phase,system was designed using MATLAB and a GUI was created, in the second phase C language code was developed for Pentium machine platform.