Engage your customers through Audio Watermark

This method is based on a controlled modification of a pre-existent audio signal. The alteration of the original signal is not detectable by human ears. The coding scheme is serial: the information bits are transmitted in a sequence, one at a time.

• Coding. The original signal has to be available in digital format to be numerically processed. It gets divided into subsequent time windows and the Short Time Fourier Transform (STFR) is computed. For each time window, the computed Fourier Transform is processed by a function in the frequency domain. Specifically, the spectrum is processed using F₁(f) if the bit to be transmitted is a 1 and using F₀(f) if the bit to be transmitted is a 0. The functions F₁(f) and F₀(f) have different shape and different parameters. Their waveform is optimized for the specific transmission channel in order to minimize the perceptibility of the alteration and maximise the information energy. When the operation is completed for all the time windows in the interval of interest, the resulting spectrum is converted back to the time domain through inverse Fourier Transform and normalised in amplitude in order to preserve the same perceived loudness. Finally, the signal is written into memory.
• Transmission. The modified audio signal can be transmitted using any audio/video transmission channel and diffused by any playback device (TV, radio, smartphone, tablet, phablet, PC, etc...). This methodology is effective also for limited band systems and in case of compressed coding.
• Receiving and decoding. The microphone embedded into the decoding device is reached by a sound pressure level that it’s turned into an electrical signal that varies with time. This signal is converted into digital domain and then processed as follows: a frequency analysis is performed on adjacent time windows. Each spectrum is analysed to check if the energy associated to the F₁(f) is higher than the energy associated to F₀(f) and it correspondingly associates a value 1 or value 0 to the relative bit. When the whole word is decoded it is compared with all the words in a list of admitted words and, in case it’s included, the decoding is considered correct.