Application of Combined Filtering in Thunder Recognition

Thunder recognition is of great interest in lightning detection and physics widely used short-range location. However, due to the complexity thunder, any single filtering method that traditional speech noise reduction technology cannot identify well thunder from complicated background noise. In this study, impact four different filters on compared, including low-pass filtering, least-mean-square adaptive spectral subtraction Wiener filtering. The original acoustic signal filtered using techniques are applied a convolutional neural network, which classified. results indicate combination filter performs best recognition. signal-to-noise ratio can be significantly improved, accuracy (93.18%) improved by 3.8–18.6% after combined method. addition, endpoints better identified frequency domain sub-band variance algorithm.