Formant tracking

The above representations are all what can be termed bottom-up representations in that they are all found from the original speech frame by means of a mathematical function or process. We can also consider other representations in which we move away from these techniqnes and move towards representations that we believe more accmately represent the properties of the signal that are most perceptually important. In Chapter 7, we showed that formants are considered one of the most important representations in traditional phonetics, bnt in addition they are used in formant synthesis, which will be 

Rgure 12.16 An example spectrogram with formants superimposed, as calculated by a formant tracking algorithm. 

In the early days of speech research, formants held the centre stage as the representation for speech. Today, however, it is extremely rare to find formants used in speech recognition, synthesis or other systems. This is partly because robust formant trackers have always proved hard to develop, so researchers moved towards cepstra and LP representations that could be robustly found from speech, even if their form didn t exphcitly show the type of information that was thought most important. Furthermore, formants are really only a powerful representation for vowel sounds they are not particularly 

---

As with formant tracking, traditional pitch detection algorithms PDAs often have three stages a pre-processing stage, where for instance the speech is downsampled or low-pass filtered, a bottom up candidate finding step, and a decision making step whereby the best candidates are picked. 

Malkin, J., Li, X., and Bilmes, J. A graphical model for formant tracking. In Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing 2005 (2005). 

---