SimNet : Simplified deep neural networks for OFDM channel estimation

• In this paper, a simplified deep neural network is proposed, which can be used for channel estimation and signal detection in OFDM system and reduce complexity. To be specific, the method of deep learning is introduced to optimize the channel estimation module of OFDM system. By building deep neural networks and training parameters at the signal-to-noise ratio of 10dB and 25dB, respectively, the channel estimation results can be optimized at a wider range of signal-to-noise ratio. In addition, the influence of training model size for channel estimation and signal detection is also researched. Compared with some other artificial intelligence aided OFDM receivers, proposed deep neural networks has shorter training time and simpler architecture. The simulation results show that by using proposed deep neural networks and training method in OFDM channel estimation, smaller mean square error and lower bit error rate can be obtained, especially in the case of clipping distortion and wide range of signal-to-noise ratio.

Ämnesord

TEKNIK OCH TEKNOLOGIER  -- Elektroteknik och elektronik -- Signalbehandling (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Electrical Engineering, Electronic Engineering, Information Engineering -- Signal Processing (hsv//eng)

TEKNIK OCH TEKNOLOGIER  -- Elektroteknik och elektronik -- Telekommunikation (hsv//swe)

ENGINEERING AND TECHNOLOGY  -- Electrical Engineering, Electronic Engineering, Information Engineering -- Telecommunications (hsv//eng)

Nyckelord

Channel estimation

Deep learning

Neural networks

OFDM

Signal detection

Bit error rate

Deep neural networks

Mean square error

Orthogonal frequency division multiplexing

Signal to noise ratio

Clipping distortion

OFDM channel estimation

OFDM receiver

OFDM systems

Training methods

Training model

Training parameters

Training time

Publikations- och innehållstyp

ref (ämneskategori)

kon (ämneskategori)