Title :
Network Optimized by K-L Transform for Digital Modulation Identification
Author :
Kuo, Yonghong ; Chen, Jian ; Xiaohua Tan
Author_Institution :
School of Communication Engineering, Xidian University
Abstract :
It is important for digital modulation identification based on neural network to determine a suitable size of a network. This paper proposes that the number of nodes in hidden layer which is the core for network optimizing can be confirmed by K-L transform. The validity and robustness are verified by simulation. The percentage of correct identification (PCI) is almost the same before and after the network optimizing.
Keywords :
Covariance matrix; Digital communication; Digital modulation; Mean square error methods; Network topology; Neural networks; Optimization methods; Redundancy; Robustness; Signal processing;
Conference_Titel :
Parallel and Distributed Computing, Applications and Technologies, 2005. PDCAT 2005. Sixth International Conference on
Print_ISBN :
0-7695-2405-2
DOI :
10.1109/PDCAT.2005.171
