In the project Spectral Envelopes in Sound Analysis and Synthesis, various methods for estimation, representation, file storage, manipulation, and application of spectral envelopes to sound synthesis were evaluated, improved, and implemented. The prototyping and testing environment VIEWENV has been developed, and a function library to handle spectral envelopes was designed and implemented.
For the estimation of spectral envelopes (chapter 3), after a definition of the requirements, the LPC, cepstrum, and discrete cepstrum methods were examined, all of which were implemented. Various separate possibilities of improvements of the discrete cepstrum method (regularization, stochastic (or probabilistic) smoothing, logarithmic frequency scaling, and adding control points to the envelope) were examined and combined. A comparison of the effects of these improvements was carried out and an evaluation with a large corpus of sound data showed the feasibility of discrete cepstrum spectral envelope estimation.
After defining the requirements for the representation of spectral envelope s (chapter 4), filter coefficients, spectral representation, geometric representations (break-point functions and splines), formant representation, and high resolution matching pursuit were examined. After a comparison of the methods in regard of the requirements, spectral representation, filter coefficients, and formant representation was chosen. A combined spectral representation with indication of the regions of formants (called ``fuzzy formants'') was defined to allow for integration of spectral envelopes with precise formant descriptions.
For file storage, data types to store each of the representations used in the project were defined in the framework of the SDIF standard sound description file format (see section 7.5).
Various types of manipulations were examined in chapter 5. Special attention has been given to interpolation between spectral envelopes, and between spectral envelopes and formants. Other manipulations, based on primitive operations on spectral envelopes, affecting the amplitude and the frequencies of spectral envelopes have been covered.
For the application of spectral envelopes to sound synthesis, the two cases of additive synthesis and filtering have been examined in chapter 6. For the latter, the conversio of the different representations to time-domain or frequency-domain filters is given.
In order to easily develop the algorithms for spectral envelope estimation, and to compare the effects of the different parameter settings, the VIEWENV spectral envelope viewing application was developed in the course of the project under the MATLAB programming environment.
Finally, the spectral envelope library, which combines all of the methods evaluated in the project, and makes them accessible to other programs, has been developed. The principles of software engineering were applied to the requirements analysis, the design of the software architecture, to the implementation, and to the testing mechanisms devised, as described in chapter 7.