Detecção de edições em áudios baseada na análise tempo-frequência e em redes neurais convolucionais

In this study, the development of an automatic splicing detection model for digital audios using convolutional neural networks was conducted. The spectrogram of the audios, calculated through different techniques: short-time fourier transform (STFT) in linear scale, STFT in mel scale, and constant Q transform (CQT), was directly provided to the network as input data. A comparative study was carried out to assess the impact of representation choice in the time-frequency domain on the model's performance in correctly classifying the original and tampered audios.