ISCA Archive Interspeech 2022
ISCA Archive Interspeech 2022

Practical Over-the-air Perceptual AcousticWatermarking

Ameya Agaskar

In this work, we demonstrate a novel technique for automatically scaling over-the-air acoustic watermarks to maximize amplitude while remaining imperceptible to human listeners. These watermarks have been demonstrated in prior work to be robust to the indoor acoustic channel. However, they require careful calibration to ensure that they are (a) detectable by the device and (b) imperceptible to humans. While previously this was done using listening tests, we show that psychoacoustic masking curves can be used to automatically scale each watermark frame's amplitude to be as high as possible while remaining below the masking level. This maximizes watermark detectability by the self-correlation decoder described in earlier work, while ensuring that the watermark is not heard.