Data can be covertly transmitted between devices using inaudible ultrasound emitted and captured by regular computer speakers and microphones.
Audio signals decompose into frequency components; ultrasound refers to frequencies above ~20 kHz which humans cannot hear.
A WebAudio-based demo implements return-to-zero frequency shift keying with a pilot tone, framing, and ASCII encoding for transmitting bits.
The prototype works at around 10 bits per second with no error correction, making it slow, interference-prone, and lacking robustness.
Developers can improve reliability by adding error correction (e.g., Reed-Solomon) or better DSP techniques like matched filtering.
Ultrasound data channels are already employed by video conferencing tools for proximity detection among nearby devices.
Get notified when new stories are published for "General AI News"