Time-variant parameters for a computer sound synthesis model are calculated by matching the instantaneous spectral centers ("brightness") and by minimizing the instantaneous spectral difference (error) between synthetic and original tones. Synthesis accuracy is gauged in terms of the average error, by a comparison of plots of synthesized time-variant spectra with those of the original tones, and by comparative listening to the tones. It was found that with the matching technique described, a nonlinear/filter model can synthesize cornonet and saxaphone tones with 4-26% error, whereas the error with simple FM synthesis is in the range of 36-72%. While the amount of computation is greater for nonlinear/filter synthesis than for simple FM (9 adds, 5 multiplies, and 4 lookups versus 5 adds, 2 multiples, and 4 lookups), the nonlinear/filter technique offers the advantages of greater accuracy and a more natural-sounding timbre synthesis over a wide range of pitch and amplitude. Data are given for synthesis of a coronet tone and an alto saxaphone tone using the nonlinear/filter technique.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.