Multiplication-free intelligent communication receiver enabled by bimodal power-aware quantization

Intelligent communication receivers based on deep neural networks typically suffer from high computational complexity due to extensive multiplication operations, which significantly limits their practical deployment in resource-constrained scenarios. To address this issue, this paper proposes a multiplication-free intelligent communication receiver by utilizing a novel Bimodal Power-Aware Quantization (BPAQ) scheme. The BPAQ scheme combines Uniform Quantization (UQ) and Power-of-Two (PoT) scaling to effectively exploit the bimodal weight distribution within the neural network, enabling the replacement of computationally expensive multiplications with efficient bit-shift operations. The proposed approach completely eliminates multiplication operations in the residual blocks of the intelligent communication receiver, thereby greatly reducing computational complexity and memory usage. Extensive experimental results demonstrate that the proposed multiplication-free receiver achieves robust Block Error Rate (BLER) performance close to that of the original unquantized receiver under low Signal-to-Noise Ratio (SNR) conditions, clearly outperforming existing quantization approaches such as uniform, binary, and ternary methods. Thus, the proposed receiver provides an efficient and practical solution for intelligent communication in scenarios with stringent computational and storage constraints.