The multiresolution Fourier transform and its application to polyphonic audio analysis

Many people listen to, or at least hear, some form of music almost every day of their lives. However, only some of the processes involved in creating the sensations and emotions evoked by the music are understood in any detail. The problem of unravelling these processes has been much less thoroughly investigated than the comparable topics of speech and image recognition; this has almost certainly been caused by the existence of a greater number of applications awaiting this knowledge. Nevertheless, the area of music perception has attracted some attention over the last few decades and there is an increasing interest in the subject largely arising from the availability of suitably powerful technology. It is becoming feasible to use such technology to construct artificial hearing devices which attempt to reproduce the functionality of the human auditory system. The construction of such devices is both a powerful method of verifying operational theories of the human auditory system and may ultimately provide a means of analysing music in more detail than man. In addition to the analytical benefits, techniques developed in this manner are readily applicable to the creative aspects of music, such as the composition of new music and musical sounds