基于电网频率的数字录音真伪鉴别研究

Authentication of digital audio recording based on power system frequency

采用电网频率鉴别数字录音真伪是司法科学领域近年来的研究热点。在北美电网频率监测系统(FNET)的基础上,构建用于数字录音信号真伪鉴别的专用标准电网频率数据库,提出一种完整的真伪鉴别算法。采用数字滤波技术和短时傅里叶变换估计蕴含在数字信号中的电网频率分量,通过补零操作及三点样条插值提高频率估计的准确度;首次考虑数字采集系统振荡器对频率数据库匹配的影响,提出一种迭代式振荡器误差校正方法;分析短时傅里叶变换的参数选择,如窗函数、补零系数对频率估计准确度的影响,最后实现了该套数字录音真伪鉴别算法。在实测数字语音信号上的实验结果表明,该方法能有效地鉴别数字信号的真伪并确定其原创时间。

Power system frequency has recently been explored for digital audio authentication in the field of forensic science. This paper firstly establishes an ad hoc power system frequency reference database according to the North American frequency monitoring system （FNET） ,and then proposes a thorough algorithm for the authentication prob- lem. Digital filtering is incorporated with short-time Fourier transform （ST^F） to estimate the coarse frequency em- bedded in a questioned digital audio recording, then zero-padding operation and three-point spline interpolation are adopted to improve the accuracy of frequency estimation. An iterative method for oscillator error correction is put for- ward for the first time considering the effect of digital acquisition system oscillator on frequency database matching. Moreover, the selection of the STF＇F parameters, such as window functions and zero-padding coefficient, are also dis- cussed. Finally, the digital audio recording authentication algorithm is implemented ; the experiment result for real test- ed digital audio recording signal indicates that the proposed algorithm can effectively realize digital audio recording signal authentication and determine its original creation time.