Blind Modulation Classification via Accelerated Deep Learning
Lingjun Zhu, Zikang Gao, Zhechen Zhu · 2019
State-of-the-art modulation classifiers often require matching channel model and perfect knowledge of channel state information to mitigate lose in classification accuracy. In this paper, we propose an accelerated deep learning algorithm to achieve the same goal without the aforementioned requirements. The proposed algorithm is able to extract the hide features in raw signal samples before proceeding to obtain the classification decision. While delivering accuracy close to the maximum likelihood classifier in AWGN channels, the resulting solution provides superior performance in fading channels with various levels of carrier phase offset.