A further application area is in watermarking - the insertion of multiple redundant, error corrected unique identification codes within audio data (such as that recorded on CD). Illicit digital copies would also carry the watermark, allowing the copy source to be identified. This would be both a deterrent and a technical solution to pirate copying of CDs.
Established methods of data insertion typically involve the substitution of either of the two least-significant bits per sample word. For digital audio systems such as CD or DAT, which store a relatively large amount of audio data (over 1400 and 1500 kbit/s in order to achieve high fidelity storage), these methods can readily accomodate concealment of data up to 48kbit/s.
In audio coding research, there has been much effort involved in psychoacoustic modelling. This identifies the regions of an audio signal which do not contribute to the human perception of the sound. In speech coding, such regions are not transmitted in order to reduce data rate.
It is proposed that a similar approach be taken to data concealment. Perceptual estimation algorithms will be developed to flag redundant areas of the waveform. Storage of hidden data in imperceptible audio regions will allow the concealment of much higher data rates. More importantly, the perceptual effect of the processing on listeners can be estimated and controlled to remain constant (unlike the established methods).
Concealed data types could include compressed speech (G.728, GSM or CELP) or even high-quality stereo audio channels coded using AC-3 or MPEG although it is not expected that these audio data compression functions be implemented during this project.
Watermark data would be a highly redundant unique serial number, probably protected with a high level of error-correction coding.
The project also includes the requirement of allowing data encryption to be applied to the secondary data, and the consideration of a variety of primary source data types (such as 16-bit audio, ADPCM-coded audio etc..). These would be not be implemented in the project but would be investigated, and if time allows, tested with the system.
This project will require a thorough investigation of available techniques and MATLAB/RLaB simulation of some of these. Once a candidate algorithm is found, the system will be coded and simulated in VHDL and implemented on an FPGA for a 16-bit audio system as shown in the figure:
In the encoding system (top path), channels of analogue audio and data to be encrypted will be input, combined and stored to DAT tape. A decoding system will mirror the encoding process, and output analogue audio and decrypted data. SPDIF (Sony/Philips Digital Interface Format) is the consumer version of the AES/EBU (IEC-958) serial data interface between DAT and CD machines.
The digital components of the system are suitable for future exploitation on asic.
Construction of such parallel systems will facilitate hardware testing, and full-system testing, using the three test modes as shown:
Normal operation involves concealment of an ancillary data stream within a primary PCM audio signal, and storage of the combined data to DAT tape for later replay and reconstruction.
Equipment and facilities used in the research:
This project is designed to lead to a Master's degree by research, and should allow a participating student to forge
strong links with the audio and broadcasting industires in Singapore.
Fig. 1 Block diagram of data concealment system as it is currently envisaged."
The project is designed to allow industrial collaboration, notably through a follow-on project involving designing an ASIC system. The ASIC system would allow integration of watermarking (anti-piracy) hardware in studio CD writers and DAT recorders, and the creation of a portable anti-piracy test station.
Fig. 2 Block diagram of portable anti-piracy test unit with data display."