WHU - Audio ENF

The Whu dataset is an audio only dataset thought for testing ENF detection. It is divided into two parts: one with recordings containing ENF traces (H1) and one without them (H0). The recordings with ENF traces are coupled with the corresponding ENF reference (H1_ref). The dataset consists of 60 real-world audio recordings captured around Wuhan University campus, featuring a diverse range of environments and conditions. The recordings were made at a sampling rate of 44.1 kHz with 16-bit quantization and mono channel. Among the 60 recordings, 50 were found to have captured and verified ENF signals, which were confirmed by comparing the recording times with a reference database. The remaining 10 recordings were made in open exterior environments with strong noise and interference, and were often affected by Doppler effects due to the user walking while recording. The final dataset has 130 audio recordings in H1 and 40 in H0, obtained by randomly cropping them to durations varying from 5 to 270 seconds. In our experiments we don’t consider H0 bacause we needed recordings coupled with ENF reference.