音频辅助的色情视频检测

long with the development of information technology, more illegal pornographic video come up to the Internet, which do harm to the society. Based on this, an efficient method to prevent it's transmission would be imperative. This paper presents a combination of GMM and HMM classification scheme to achieve the pornographic video content detection based on audio. First, Soundtrack extracted from the video stream are sub-framed and extracted features, followed with a pre-trained GMM model to classify the audio stream and keep only two types sounds frame, music and music&speech; third, use pre-trained HMM model to determine whether they are pornographic video. The result confirm its validity.