2. Where is it used?
Military Applications
Intercepting
Jamming
Civilian Applications
Interference
identification
Spectrum
management
3. Drawbacks of previously established methods
• Depends of accuracy of operator
• Very slow to detect
Operator
Monitored
• Lot of simultaneously active H/W
• Again dependent on operators inference
Bank of
demodulators
5. Methods for modulation recognition
Decision
Theory based
• Decision tree
flowchart
Likelihood
based
• Maximum
Likelihood-
based
Machine
Learning
• Artificial Neural
Network
6. Methodology
Preprocessing
• Signal Isolation
• Signal Segmentation
Key Feature
Extraction
• γmax , σap , σdp , P, σaa , σaf ,
σa , µa
4,2 , µf
4,2
Modulation
classification
• Decision tree
• ANN
9. Summary of features and typical thresholds
Feature
Distinction
Thresholds for
(30dB SNR)Subset 1 (High) Subset 2 (Low)
Ratio P USB, LSB Others 0.75
Sigma dp MFSK, DSB, SSB AM, MASK 0.5
Sigma ap FM, MFSK, DSB MASK, MPSK 4
Gamma Max MASK , AM , DSB MFSK, FM ,MPSK 10
Mue a AM MASK 1.526
Mue f FM MFSK 1.6
Sigma aa 4ASK 2ASK 0.2
Sigma a 4PSK 2PSK 0.04
Sigma af 4FSK 2FSK 0.8
10. Demerits of feature extraction based approach:
• Threshold values is dependent on fc/fm ratio, signal to noise ratio,
modulation index.
• Frequent operator intervention for recalibrating thresholds.
• Holds good for high values of SNR
11. Introduction to Artificial Neural Networks (ANN)
• Artificial neural network (ANN) is a machine learning approach that models
human brain and consists of a number of artificial neurons
• An Artificial Neural Network is specified by:
− neuron model: the information processing unit of the NN
− an architecture: a set of neurons and weighted links connecting neurons along with biases
− a learning algorithm: used for training the NN by modifying the weights in order to model a
particular learning task correctly on the training examples.
12. Need for Neural network approach
Speed
• Does not contain complex real-time processing
Programming
• No need to manually program thresholds, it learns from examples
Hardware
• Post-training, it is just a combinational circuit
15. No. Modulation
Type
Lowest SNR for successful
detection
1 AM 1 dB
2 DSBSC 0 dB
3 USB 2 dB
4 LSB -3 dB
5 FM 1 dB
6 BASK 11 dB
7 BFSK -2 dB
8 BPSK -1 dB
9 4 – ASK -6 dB
10 4 – FSK 5 dB
11 4 - PSK -11 dB