DAFX: Digital Audio Effects, Second Edition

8.4 Conclusion

The central topic of this chapter is the division of the audio signal into its source signal and a time-varying filter derived from the spectral envelope of the signal. These two features are individually processed before synthesis of an output signal. The source-filter model of an audio signal, originally a basic technique for speech processing, allows the implementation of several digital audio effects based on these two global features of an audio signal and opens up new vistas for experimentation and further research. These global features can either be extracted by time-frequency techniques (FFT/IFFT) and the cepstrum method or time-domain techniques based on linear prediction (LPC). Both techniques deliver a source-filter model of the audio input signal. Beyond it, they allow the extraction of further global features such as pitch or fundamental frequency, and this will be further described in Chapter 9. A further alternative to the source-filter processing presented in this chapter, is the separation of the audio signal into individual components such as sinusoids and noise, which is discussed in Chapter 10.

1 With the expectation value E{ · }.

2 The multiplication of the expectation values by the block length N does not have any effect on the normal equations.