Audio Watermarking

A course project undertaken as a part of Digital Signal Processing course, April 2007.
 
Abstract: An audio watermarking system is developed for embedding ascii/binary data into CD quality audio files. The watermark is added to short (1024 samples) frames of the audio signal. The watermark signal is based on a psychoacoustic model. A psychoacoustic model was heuristically developed based on the Fletcher-Munson curves. A unit power watermark signal of a fixed frame length (1024 samples) is obtained based on this non-adaptive psychoacoutic model. The ascii/binary data are converted into a bit stream. For enhanced security, the bit stream is then modulated with a PN sequence to obtain a direct sequence spread spectrum (DSSS) sequence. The bits of the DSSS sequence are encoded in the successive frames of the audio signal, so that each frame of the audio signal has one bit encoded. The power of the watermark signal in each frame is scaled down to be 20 dB below the signal power in the frame. The watermark signal is added to the frame if the bit to be encoded is 1 and the watermark sign is reversed to encode 0. At the decoder, each frame is decoded using a filter matched to the watermark signal. The DSSS bit sequence is demodulated using the known PN sequence used at the encoder. Upto 44 bits of data can be encoded per second of audio signal, depending on the length of the PN sequence. The system produces near transparent quality watermarked audio signal since it is based on a psychoacoustic model. The system can be used to embed metadata such as the file, album and artist information. The length of the watermark sequence which can be encoded depends on the length of the audio signal. Since DSSS is used for a better security, there is also an overhead caused by the PN sequence. The watermarking system is basic in its functionality and is not robust to filtering, noise, and compression. The system was implemented in C and is capable of watermarking .wav audio files.
 
Shown here is a sample audio signal and the corresponding watermarked audio signal. (Please use the link above the corresponding file to download the .wav files in case you are not able to play them on the browser).
 
 
Comments