Coal gangue recognition in the strong background noise using two-level auditory feature fusion with attention mechanism

• An auditory model is introduced into the coal gangue recognition task. • Two-level auditory features were extracted, which better express the information. • A Fusion recognition model based on attention mechanism was constructed. • Experimentation results show that the proposed method has superiority in filtering noise. In the complex environment of top-coal mining, the recognition of collapsed coal and gangue is significantly impacted by strong background noise. To improve the noise resistance of coal gangue recognition, this paper proposes an anti-noise recognition method that utilizes two-level auditory feature fusion and attention mechanism. The proposed method effectively captures the deep-level information of caving coal and gangue. Thereafter, the signals are processed through feature extraction and attention mechanism fusion, which can more accurately represent coal and gangue sound information from different levels. Compared with other conventional methods, results demonstrate that the proposed method achieves over 92 % recognition accuracy under all signal-to-noise ratio ( SNR ) conditions. The reduction in coal gangue recognition performance is limited to 5.98 %, especially when the SNR of –5 dB, compared to 21.20 % for the time spectrogram method, indicating the method’s superiority in noise filtering. This study provides a new and effective means for coal gangue recognition.