(英) |
Digital audio watermark is a technique for hiding information in audio data so that it cannot be recognized by humans.Audio data is often used as evidence such as a trial.
Audio data is easily to edited by software, so that audio expert opinion by sound spectrogram analysis is performed to enhance its authenticity.It is, however, due to the development of artificial intelligence and machine learning in recent years, its editing quality has been significantly improved and it is difficult to easily distinguish whether editing is done or not. Therefore, a decline in the evidential value of the audio data is regarded as anxiety. In this report, we propose a method to tamper detection of audio data and specifying tamper place without requiring comparison with original data and spectrogram analysis.By dividing the data into blocks on the time axis and embedding the flag in the amplitude histogram for each block, it is possible to identify the tampered data by the presence or absence of the flag.Several numerical experiments show the effectiveness of this method. |