关键词: 关键词: 关键词: 多媒体取证, 手机来源识别, 频谱融合特征, 特征选择, 平均识别准确率

Abstract: With the popularity of cell-phone recording devices and the availability of various powerful and easy to operate digital media editing software, source cell-phone identification has become a hot topic in multimedia forensics, a cell-phone source recognition algorithm based on spectral fusion features is proposed to solve this problem . First, the same speech spectrogram of the different cell-phone is analyzed, found that the speech spectrum characteristics of different cell-phone is different; then the speech spectral logarithmic spectrum, phase spectrum and information quantity characteristics are researched; Secondly, three features are connected in series to form the original fusion feature, and the sample feature space is constructed with the original fusion feature of each sample; finally, the CfsSubsetEval evaluation function of WEKA platform is selected according to the best priority search method to select feature, and LibSVM is used to model training and recognition of the sample after feature selection. Twenty-three popular models of the cell-phone are evaluated in the experiment, the results show that the proposed spectral fusion feature has better identification accuracy in cell-phone brands than spectral single feature and the average recognition rates achieved 99.96% and 99.91% on the TIMIT database and CKC-SD database. In addition, it is compared with the source identification algorithm of Hanilci based on Mel frequency cepstral coefficients, the average recognition accuracy was improved by 6.58% and 5.14%. Therefore, the proposed algorithm can improve the average recognition accuracy of cell-phone source identification, and effectively reduce the false positives rate of cell-phone identification.

Key words: audio forensics, source cell-phone identification, spectral fusion feature, feature selection, average recognition accuracy