Formant values

We know fi om acoustic phonetics that typical formant values for an /ih/ vowel are 300Hz, 2200Hz, and 3000Hz. Formant bandwidths are harder to measure accurately, but less us assume a value of 250Hz for all three formants. Assuming a sampling frequency of Fs = 16000/7z, the following table shows how to calculate the poles from the formant frequencies and bandwidths. 

These equations show the main advantage of the formant technique, namely that for a given set of formant values, we can easily create a single transfer function and difference equation for the whole oral tract. In a similar fashion, a nasal system can be created, which likewise links the values of the nasal formant to a transfer function and difference equation. 

When generating speech, in principle all the parameters (that is, the formant values, bandwidth, voicing, FO etc) could be changed on a sample by sample basis. In principle however, a slower rate of parameter update, such as every 5ms, is used. 

Figure 13.5 Example spectrogram with formant values marked...

Ease of data acquisition Whether the system is rule-driven or data-driven , some data has to be acquired, even if this is just to help the rule-writer determine appropriate values for the rules. Here linear prediction clearly wins, because its parameters can easily be determined from any real speech w aveform. When formant synthesisers were mainly being developed, no fully reliable formant trackers existed, so the formant values had to be determined either manually or semi-manually. While better formant traekers now exist, many other parameters required in formant S5mthesis (e.g. zero loeations or bandwidth values) are still somewhat difficult to determine. Articulatory synthesis is partieularly interesting in that in the past it was next to impossible to acquire data. Now, various techniques such as EMA and MRI have made this much easier, and so it should be possible to collect much bigger databases for this purpose. The inability to collect accurate articulatory data is certainly one of the main reasons why articulatory synthesis never really took off. 

On the other hand, with formant synthesis, we can in some sense relate the parameters to the phonetics in that we know for instance that the typical formant values for an /iy/ vowel are 300Hz,... 

On the other hand, with formant synthesis, we can in some sense relate the parameters to the phonetics in that we know for instance that the typical formant values for an /iy/ vowel are 300 Hz, 2200 Hz and 3000 Hz. However, because we can t find the required parameters easily for some arbitrary section of speech, we have to resort to a top-down frilly specified model for transitions, which is often found severely lacking when compared with real transitions. Because parameter collection is so hard, it is difficult to come up with more-accurate phonetics-to-parameter models since the development-and-testing cycle is slow and inaccurate. 

---