Frequency domain digital encoding technique for musical signals

Abstract

An apparatus for encoding and decoding musical signals in digital form employs conversion to the frequency domain. Discrete Fourier transform coefficients are calculated from time domain digital data at selected frequencies which are organized in sets of octavely related frequencies. Selecting frequencies in this manner greatly reduces the data rate by eliminating coefficients for many higher frequencies without a great loss of fidelity because the provision of a predetermined number of frequencies in each octave approximates the tonal response of the human ear. A further refinement is selection of frequencies to correspond to the frequencies of musical notes where the greatest energy can be expected. The complex number discrete Fourier transform coefficients can be converted to polar form with logarithmic magnitude values resulting in a further data rate reduction because the human ear is responsive only to logarithmic amplitude and because the human ear is relatively insensitive to phase angle enabling fewer bits to be used to represent the angular part without loss of fidelity. A digital filter can be easily implemented by manipulation of the magnitude part of the discrete Fourier transform coefficients.