Electrical network frequency (ENF) has been used as evidence for location forensic. To determine location, we need accurate ENF information from noisy media files, select features of the ENF signal and the classify it based on previous knowledge of different grids.

In this report, we first utilize multiple harmonics in frequency domain and an error correction method for accurate ENF estimation. Utilizing multiple harmonics faces the challenges of obtaining accurate signal to noise power ratio for weighted combine, we propose to use signal level to noise level as weights. Moreover, we exploit the fact that the ENF signal doesn’t jump back and forth within a short period to design an error correction method, which is capable to get rid of inaccurate ENF estimation.

Second, we propose to combine the features both in time domain and frequency for grid classification. Based on the power signal characteristics, we propose to use crest factor and the ratio of first harmonic to the third harmonic to evaluate the distortion of power waveform. Moreover, we propose to categorize the signals based on nominal frequency and source type, and design proper features for each category via cross validation method. Our proposed classification scheme can achieve an accuracy rate of 94% for practice dataset.

Third, we design and build a small yet effective sensing circuit with accurate timing. We record 10 hours of data from both home, university in different time. Using our proposed classifier, we learnt that with confidence level of 47.6%, the waterloo grid is part of grid C. Since the confidence level is low, we consider waterloo grid belongs to ”None of Above Options”.