A technique for the analysis, synthesis, and modification of musical tones is presented which uses an overlap-add sinusoidal model of the digitally recorded tone. Automatic analysis of model parameters is achieved using a novel analysis-by-synthesis procedure that incorporates an iterative vector approximation technique. This analysis procedure is described in detail, and an equivalent frequency-domain algorithm introduced which takes advantage of the computational efficiency of the fast Fourier transform (FFT) algorithm. Furthermore, a refined overlap-add sinusoidal model is derived that is capable of synthesizing and modifying quasi-harmonic tones without artifacts. Analysis-by-synthesis overcomes problems encountered in the digital phase vocoder due to analysis of the attack and release portions of tones, is well suited to analyzing inharmonic and pitch-varying tones, and provides greater accuracy than spectral sampling approaches currently used for sinusoidal modeling. Overlap-add sinusoidal modeling maintains the generality and flexibility of additive synthesis, and may be implemented much more efficiently than classical additive synthesis by using the inverse FFT algorithm. The analysis-by-synthesis/overlap-add sinusoidal model has been successfully applied to the analysis, synthesis, and modification of instruments from the brass, woodwind, and string families as well as the piano. The authors plan to further investigate the use of this model to modify inharmonic tones.
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.