DAFX: Digital Audio Effects, Second Edition

10.5 Conclusions

In this chapter, we have shown how the use of spectral models based on a sinusoidal plus residual decomposition can lead to new and interesting sound effects and transformations. We have also seen that it is not easy nor immediate to get a good spectral representation of a sound, so the use of this kind of approach needs to be carefully considered, bearing in mind the application and the type of sounds we want to process.

For example, most of the techniques here presented work well only on monophonic sounds and some rely on the pseudo-harmonicity of the input signal.

Nevertheless, the use of spectral models for musical processing has not been around too long and it has already proven useful for many applications, such as the ones presented in this chapter. Under many circumstances, spectral models based on the sinusoidal plus residual decomposition offer much more flexibility and processing capabilities than more immediate representations of the sound signal.

In general, higher-level sound representations offer more flexibility at the cost of a more complex and time-consuming analysis process. It is important to remember that the model of the sound we choose will surely have a great effect on the kind of transformations we will be able to achieve and on the complexity and efficiency of our implementation. Hopefully, the reading of this chapter, and the book as a whole, will guide the reader in taking the right decision in order to get the desired results.