Экспериментальные исследования эквалайзера на основе нейронных сетей с обучением в многолучевом радиоканале

Abstract

The frequency-selective fading inherent in multipath channels leads to the inter-symbol interference, which affects the data transmission reliability. Implementation of the equalizer in data transmission system exploits the inter-symbol interference thus increasing the energy efficiency at the receiver. The existing decision feedback equalizer (DFE) and maximum-likelihood sequence estimator (MLSE) have following disadvantages: DFE has poor energy efficiency and MLSE is computationally complex. The neural network equalizer algorithm proposed in this paper has proved to have better performance in terms of the energy efficiency than DFE, while having the similar computational complexity. The neural network equalizer was implemented in the data transmission system along with carrier recovery, clock recovery, automated gain control, matched filtering, interleaving, de-interleaving, turbo-coding, turbo-decoding blocks. The energy efficiency of the neural network equalizer was experimentally estimated in different conditions: 1) single-path flat radio frequency channel with AWGN. The effects of carrier frequency offset between the transmitter and the receiver (up to 5kHz) and the Doppler shift (up to 150 Hz) were investigated; 2) stationary multipath channel environment. The performance of the equalizer with different number of iterations was measured; 3) non-stationary fast-varying multipath channel. The impact of the step size in learning procedure of the equalizer on the performance was investigated. In addition, the combination of the proposed neural network equalizer and turbo-code was investigated in all the mentioned channels in order to test the equalizer’s performance in data transmission system.

• equalizer
• multipath channel
• neural networks
• intersymbol interference

Аннотация (на русском)

Экспериментальные исследования эквалайзера на основе нейронных сетей с обучением в многолучевом радиоканале

По причине присутствия частотно-селективных замираний в многолучевом радиоканале, при передаче данных возникает межсимвольная интерференция, которая может существенно понижать энергетическую эффективность систем связи. Использование эквалайзера в системе приема позволяет уменьшить межсимвольную интерференцию, повысив энергетическую эффективность. В данной работе в экспериментальном макете реализована конфигурация эквалайзера на основе нейронных сетей, осуществлена передача и прием в условиях многолучевого распространения радиоволн с частотно-селективными и изменяющимися во времени замираниями, исследована работа эквалайзера в данных условиях.

For citation

References

1. H.С.Myburgh, J.C.Olivier. Near-optimal low complexity MLSE equalization. // Wireless Communications and Networking Conference, 2008, IEEE, pp.226-230
2. H.С.Myburgh, J.C.Olivier. Low complexity MLSE equalization in highly dispersive Rayleigh fading channels. // EURASIP Journal on Advances in Signal Processing. Dec.2010 p. 10.
3. Д.Р.Валиуллин, П.Н.Захаров. Эквалайзер на основе нейронных сетей с обучением в многолучевом канале. // Успехи современной радиоэлектроники, 2016, №11, стр. 200-202