论文部分内容阅读
根据数字录音设备在录音过程中不仅记录语音内容本身,还携带微弱的电网信号的特点,提出一种基于电网频率的特征提取和复制粘贴篡改检测方法。方法利用Duffing带通滤波器获取携带电网信号的语音信号,然后提取该语音信号的Mel频标倒谱系数(MFCC),最后将特征参数运用于基于量子模距离的判决模型,对音频信号进行复制粘贴篡改检测。通过实验分析,该方法在数字语音删除、复制粘贴篡改检测方面,具有较高的正确率;音频信号越长,检测的正确率越高;MFCC参数提取过程中,帧长和帧移越小,检测的正确率越高。该方法可为音频信号篡改检测提供一个新的研究方向。
According to the digital recording equipment not only recording the voice content itself, but also carrying the weak signal of the power grid, a method of feature extraction and copy-and-paste tampering detection based on grid frequency is proposed. The method uses Duffing band-pass filter to obtain the voice signal carrying the power network signal, and then extracts the Mel-Frequency Cepstral Coefficient (MFCC) of the voice signal. Finally, the characteristic parameters are applied to the decision model based on quantum distance to copy the audio signal Paste tampering detection. Through the experimental analysis, this method has a higher correct rate in the digital voice deletion, copy and paste tamper detection; the longer the audio signal, the higher the detection accuracy; the smaller the frame length and the frame shift in the MFCC parameter extraction process, The higher the detection accuracy. This method can provide a new research direction for tampering detection of audio signals.