A deep learning-based efficient MIMO signal detection model for mining environments is proposed to address the low detection efficiency issue of intelligent receivers. The model consists of a nonlinear mapping network and an error correction network. The nonlinear mapping network is responsible for the initial recovery of received signals into binary bit signals, while the error correction network corrects the errors introduced by the nonlinear mapping network, thereby improving the signal detection accuracy. The performance of the proposed model is validated through simulations conducted in a mining MIMO communication system. The results demonstrate that the model outperforms traditional receivers in mining MIMO communication environments, especially when the modulation scheme at the transmitter, channel coding method, and channel environment change. Additionally, compared to deep receiver models, the proposed model achieves higher detection efficiency. This research presents a novel solution to the low decoding efficiency problem of intelligent receivers and provides evidence of the superior.