Recognition method of coal mine gas and coal dust explosion based on sound spectrogram and SVM

In order to improve the detection accuracy of coal mine gas and coal dust explosion, recognition method of coal mine gas and coal dust explosion based on the sound spectrogram and SVM was proposed: installing mining pickups in key monitoring areas of underground coal mines, monitoring of working sounds of underground coal mine equipment and environmental sound in real-time. The collected sound was extracted from the sound spectrogram composed of MFCC, and the energy, entropy, contrast, and correlation of 0°, 45°, 90°, and 135° were obtained by calculating the gray co-generation matrix of the sound spectrogram. Calculated their mean and standard deviation as the texture features of the sound spectrogram image, which constituted the feature quantity of the sound, which brought to SVM to establish the sound recognition model of coal mine gas and coal dust explosion. For the sound to be tested, the texture features were also extracted and input to the trained recognition model for sound recognition and classification, which has passed the test verification. Firstly, the MFCC feature values of different sounds, such as mining equipment operation, gas and coal dust explosion were extracted, and the distribution of MFCC feature values of different sounds was analyzed; the sound spectrograms of different sounds were extracted, and the mean and standard deviation of the sound spectrograms of different sounds were analyzed by analyzing the feature parameters of energy, entropy, contrast, and correlation, and the grayscale co-generation matrix feature parameters of the sound spectrograms constitute the feature quantities that can effectively characterize the sound signals. Secondly, the sound to be measured was input into the recognition model established to complete the recognition and classification. The results show that the recognition rate of the proposed method reaches 95%, and the overall recognition performance is higher than others models. Finally, the experimental results of Bayesian parameter optimization show that the recall rate and recognition rate of the optimized SVM recognition model increased by 10% and 3% respectively. It was better than the recognition model before optimization, which can meet the needs of coal mine gas and coal dust explosion sensing and alarming.