Research on Pornography Detection in Speech Recognition Based on Convolutional Neural Networks and Design of an Anti-Pornography System

For the classification and recognition of pornographic audio, the accuracy ranges from 66.81 % to 95.74 %. Based on this study, the paper proposes a speech recognition pornography detection system that utilizes convolutional neural networks. This system ensures that detection can be completed with a reduced sample vocabulary library. Moreover, the detection method based on periodic pornographic feature audio offers a novel approach to pornography detection, which holds significant value for the research and application of pornographic audio detection systems.