TOA Estimation Scheme Based on CNN for B-IFDM-Based Preambles

To satisfy the requirement of power spectral density and occupied bandwidth, uplink transmission is based on block interleaved frequency division multiplexing (B-IFDM) structure on unlicensed spectrum. However, the accuracy of time of arrival (TOA) estimation is damaged due to false peaks of autocorrelation for the reference signal of B-IFDM structure, when the existing TOA estimation scheme is used. In this paper, we propose a new TOA estimation scheme based on convolutional neural network (CNN). The proposed scheme overcomes the negative effect of false peaks and greatly outperforms the existing scheme.