Efficient Beamforming Training and Channel Estimation for Millimeter Wave OFDM Systems

We study the problem of downlink beamforming training and channel estimation for millimeter wave (mmWave) OFDM systems, where a hybrid analog and digital beamforming structure is employed at the transmitter (i.e., base station) and an omni-directional antenna or an antenna array is used at the receiver (i.e., user). To efficiently probe the channel, we form multiple directional beams simultaneously at the transmitter and steer them towards different directions. The objective is to devise the beam training sequence and develop an efficient algorithm to estimate the channel. By exploiting the sparse scattering nature of mmWave channels, the above problem is formulated as one of sparse encoding and signal recovery, which involves finding a sparse sensing matrix to compress the sparse channel and an efficient channel estimation algorithm to recover the sparse channel from compressive measurements. In this article, we propose a sparse bipartite graph code-based algorithm, where a set of bipartite graphs are employed to encode the sparse channel and a simple decoding procedure that relies on the presence of a No-Multiton-graph (NM-graph) is used to reconstruct the sparse channel. Theoretical analysis shows that our proposed method can help achieve a substantial training overhead reduction. Simulations are provided to show the effectiveness of the proposed algorithm and its performance advantage over compressed sensing-based methods.

I. Introduction

Millimeter wave (mmWave) communication is regarded as a promising technology for future cellular networks due to its large available bandwidth and the potential to offer gigabits-per-second communication data rates [1]–[3]. Utilizing large antenna arrays is essential for mmWave systems as mmWave signals incur a much higher free-space path loss compared to microwave signals below 6 GHz [4], [5]. Due to the use of a large number of antennas, the channel matrix/vector to be estimated has a large size, which makes channel estimation a challenging issue. In addition, hybrid analog-digital precoding structures are usually employed in mmWave systems to reduce the hardware complexity and power consumption [6]–[9], which causes the so-called channel subspace sampling limitation issue and makes it even harder to acquire useful channel state information (CSI) during a practical channel coherence time [10].