ModulationClassification_presentation_slides.pptx

Modulation classification introduction

1. Modulation Classification

2. Motivation “The goal is to determine the modulation scheme or technique used to encode information in a received signal.” Signal Identification: How information is coded? Adaptive Communication : Classify modulation Scheme and adapts demodulator. Interference and Security: Distinguish between legitimate and unwanted signals thereby it is important in signal intelligence and electronic warfare . Spectrum Management: It helps in monitor and manage the allocation of radio frequency spectrum.

3. Importance Military It is necessary to extract the information after discovering the enemy's signal. Prerequisite for jamming, eavesdropping, and intercepting enemy communications Application of electronic countermeasure Civil Prevent the spectrum from being maliciously abused Distinguish the signal modulation types of each frequency band to ensure the rational use of spectrum resources Restricts unauthorized users use illegal radio stations

4. Modulation classification Algorithms Based on Likelihood Decision Theory Average Likelihood ratio test (ALRT) Generalized Likelihood ratio test (GLRT) Hybrid Likelihood ratio test (HLRT) Based on feature extraction Instantaneous attribute Constellation Higher order Cumulant Cyclic Spectru m Based on deep learning Recurrent Neural Network Convolutional Neural Network

5. FIGURE Automatic Modulation Classification (AMC) techniques ML : Maximum likelihood classifier ALRT : Average Likelihood Ratio Test GLRT : Generalized Likelihood Ratio Test HLRT : Hybrid Likelihood Ratio Test MFCCs : Mel-Frequency Cepstral Coefficients : Maximum value of normalized-centered instantaneous amplitude's power spectral density : Standard deviation of the absolute value of the instantaneous phase's nonlinear component : Standard deviation of the direct value of the instantaneous phase's nonlinear component P : Spectrum symmetry : Standard deviation of the normalized-centered of the instantaneous amplitude's absolute value : Standard deviation of the normalized instantaneous frequency's absolute value : Standard deviation of the normalized-centered instantaneous amplitude : Kurtosis of the normalized instantaneous amplitude : Kurtosis of the normalized instantaneous frequency  

6. Likelihood based classifier Likelihood calculation Likelihood comparison Joint Likelihood function

7. ALRT Observed waveform r(t)=s(t)+w(t), 0 < t < NTs classifying between K possible modulations, by observing N phase-rotated and scaled symbols in AWGN, can be stated as the following multiple hypothesis testing problem: The unknown parameters ( data symbols and carrier phase )are modeled as random variables and their joint pdf is either known or can be hypothesized, the "optimal" solution is the ALRT

8. Disadvantages ALRT is more complex in Comparison to ML. Pdf of unknown channel parameter are require to be known accurately.

9. GLRT When it is difficult to attach a pdf to parameter of interest, therefore modelled as fixed but unknown quantities Disadvantages ML estimation requires a nonlinear, multidimensional maximization procedure Complexity is decreased but it becomes biased to higher order modulation in case of nested modulation scheme.

10. HLRT Combination of ALRT and GLRT Here, some of the unknown parameters in p as random (p2)and the rest as deterministic but unknown variables ( p1).