CHANNEL ESTIMATION AND DETECTION OF FILTER BANK MULTICARRIER SYSTEM BASED ON RESNET-DNN

Filter bank multicarrier FBMC systems are widely concerned because of their high spectral flexibility and efficiency. Two channel estimation and detection schemes of FBMC system based on residual neural network are proposed for the inherent imaginary interference of FBMC system. In scheme 1, residual neural network was used to model the channel estimation module to complete the approximation of the time-frequency response matrix of sparse channel to the time-frequency response of real channel. In scheme 2, residual neural network was used to model and integrate channel estimation, channel equalization, OQAM demodulation and decision module. Experimental results show that the two schemes have better bit error rate performance than the traditional channel estimation algorithms.