Table 7.

Comparison of Technique 1 and Technique 2 for modulation scheme classification and parameter optimization.

Aspect	Technique 1	Technique 2
Methodology	Hierarchical classification with primary and parameter-specific classifiers	Direct classification of full dataset without hierarchy
Primary classification accuracy	98.18% (decision tree)	N/A
FSK, PSK M classification	99.81% (decision tree)	N/A
CPM M classification	99.81% (ensemble subspace KNN)	N/A
CPM L classification	99.84% (ensemble subspace KNN)	N/A
CPM h classification	99.91% (ensemble subspace KNN)	N/A
CPM mode classification	99.3% (narrow neural network)	N/A
PSK/FSK mode classification	99.8% (decision tree)	N/A
Overall mode classification accuracy	99.41% (narrow neural network and decision tree)	98.35% (bilayered neural network) for direct classification
Classifier types	Decision tree, narrow neural network and ensemble subspace KNN	Neural network
Approach	Hierarchical classification by modulation type and parameters (M, h, L, mode)	Single-step mode classification