A novel modulation classification method in cognitive radios using higher-order cumulants and denoising stacked sparse autoencoder

In this paper, we propose a novel modulation classification method based on deep network as well as higher-order cumulants. The proposed algorithm uses the higher-order cumulants as the features, and thus achieves impressive noise suppression. We use Stacked Denoising Sparse Autoencoder as a classif...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2016 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA) s. 1 - 5
Hlavní autoři: Xu Zhu, Fujii, Takeo
Médium: Konferenční příspěvek
Jazyk:angličtina
japonština
Vydáno: Asia Pacific Signal and Information Processing Association 01.12.2016
Témata:
Cognitive radio
Cost function
Feature extraction
Manganese
Modulation
Noise reduction
Training
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:In this paper, we propose a novel modulation classification method based on deep network as well as higher-order cumulants. The proposed algorithm uses the higher-order cumulants as the features, and thus achieves impressive noise suppression. We use Stacked Denoising Sparse Autoencoder as a classifier for single-carrier modulation classification. This classifier can classify different modulated signals by cumulants automatically, and omit the decision of feature thresholds. A very different aspect from conventional neural network is its stacked structure, which simplifies an exponentially large number of hidden units by a multi-layer construction. Moreover, the better performance of backpropagation and network tune can be achieved while using Stacked Sparse Autoencoder. In addition, Denoising process improves the performance of noise suppression by training the network with a corrupted database. The performance of the multi-classes classification is given by simulations, which indicates that there is a significant performance advantage over the conventional methods.
DOI:10.1109/APSIPA.2016.7820860