Non-invasive Identification of Audio Content for High Resolution Applications

A non-invasive method of audio file identification is described to ascertain the proximity of processed audio, of possibly dubious origin, to a reference file. A differential correlation function forms the basis of a comparative metric. Testing embraces linear and non-linear processes including perceptual based codecs. Applications include DVD_A and SACD when used as the only legitimate non-watermarked release formats. 1 INTRODUCTION There is a requirement to be able to detect the authenticity of recorded information especially where audio material has been illegally copied. One technique is to embed a watermark [1] in order to carry data useful for stock control and other financial transactions where ownership must be demonstrated. This carries a number of potential problems where audibility is of particular significance in high-quality applications. However, it is recognised that for network delivery of audio content that does not necessarily have to be of the highest quality, the concerns of watermarking are less controversial especially where the watermark must be robust.