Research on the blind equalization technology based on the complex BP neural network with tunable activation functions

The blind equalization technology was proposed to reduce the inter-symbol interference (ISI) without training signal sequences. It saved the bandwidth resources and increased the correctness of signal transmissions. The blind equalizers had consequently been an indispensable part of the communication system. The blind equalization technology based on backward propagation (BP) neural networks had been widely studied with a variety of developments and improvements. In this paper, a new equalizer based on the complex BP neural networks with tunable activation functions (TAF) is proposed, algorithms were derived and simulations were done. Results showed that the proposed equalizer would work faster with less square errors than that based on traditional BP neural networks. It worked very well with zero byte-error rates (BER) when signal-to-noise (SNR) of the communication system was larger than 15 dB.