Data Transmission Based on RNN Compensation over Mobile Voice Channel
Xiang Gao, Tao Peng · 2023
Data transmission over mobile voice channel (DoV) suffers from vocoder compression and smartphone de-noising. Most previous algorithms ignore these two factors leading to the generated signal being distorted severely during real network transmission. This paper proposes a data modulation algorithm that maps data on the amplitude of different frequency components, and combined with interpolation 0 in frequency domain to obtain short-time stability. Then, a recurrent neural network (RNN) based compensation algorithm is proposed to compensate for distorted signal. As a result, this paper achieves 3.1kb/s data rate with an average 4.5% bit error rate (BER) in real network experiments. Numerical results suggest that the proposed algorithm has a considerable advantage in resisting smartphone de-noising and is capable of reducing distortion generated during transmission.