Extraction and Filtering of Electric Network Frequency Using Improved Matrix Pencil and Quadratic Box Plot-Empirical Wavelet Transform

Abstract

The extraction and filtering of electric network frequency (ENF) is significant for verifying the authenticity of digital audio. However, there are many challenges in accurately extracting ENF from digital audio, which makes it difficult to establish an effective matching relationship with the database. To address this problem, an improved matrix pencil (IMP) method is presented to extract ENF signals for phase measuring units. The power grid signal is constructed into a Hankel matrix, which is decomposed into singular values and filtered out the harmonics of the power grid using an adaptive order determination method. By embedding ENF as a watermark into digital audio through encryption technology, a quadratic box plot (QBP) is proposed to detect potential outliers caused by the bit error rate. Next, the empirical wavelet transform (EWT) is used to filter out Gaussian white noise between different power equipment to improve the similarity of database matching. Integrating the IMP and QBP-EWT, examples from the dataset demonstrate that the proposed ENF extraction and filtering framework has a higher assessment performance. Compared with several commonly used methods, our framework has profound outlier identification ability and effectively improves the accuracy of database matching.