Research on Voiceprint Recognition Method of Hydropower Unit Online-State Based on Mel-Spectrum and CNN

In order to realize the non-stop voiceprint detection of the hydropower unit, a voiceprint recognition model of the hydropower unit based on the Mel-spectrum and convolutional neural network (CNN) is proposed. First, voiceprint information is collected by bone conduction voiceprint sensor, and is preprocessed by Mel time-frequency diagram to reduce dimensionality. Then, the Mel-spectrum is used as the input of the CNN network for deep learning, and an online monitoring and fault pre-diagnosis model of hydropower unit based on voiceprint recognition technology is developed. Through hyperparameter adjustment and network structure optimization, the accurate identification of voiceprint signals in different operating conditions is realized. The results show that the Mel-spectra-CNN recognition model proposed in this paper achieves 98.79% recognition rate of acoustic signals in 4 different working conditions, which improves the accuracy of fault diagnosis and the safety of hydropower units.