I need to do phonetic analysis of several dozen brief tokens. Specifically, I want to extract FFT/LPC profiles of fricatives, and would like to experiment with different algorithms for creating profiles and comparing them across tokens. Which package would be easiest to use? Praat is obviously very phonetics-oriented, but its scripting language seems less well-developed than that of R. R has packages for everything, but I don't know how good are its packages for speech analysis and display. There's also Octave, which I have never used, and python. I would like the process to be simple and easy to figure out.
1 Answer
In Phonetics, there are three steps needed in order to process and analyse your data: segmenting(/labelling), numerical analysis and data visualisation.
Each of these steps can be carried out independently, therefore you can choose a distinct tool for each of them. But the interoperability is rarely possible, consequently the combination are limited.
Here is a website summarising tools that exist: http://liceu.uab.es/~joaquim/phonetics/fon_anal_acus/herram_anal_acus.html
If you are only interested in numerical analysis and data visualisation with R because your corpus is already segmented, look at the CRAN repository (keywords: phonetic, sound): https://cran.r-project.org/web/packages/available_packages_by_name.html
These packages are not good or bad, they just don't offer all the options. Then, you have to combine them, even to combine with your own code.
The easier way for you is to opt for tools that you know already. If R is a language with whom you are used to, it is better to continue to work with this language.
-
Thanks for the clear explanation. The segmentation would be done in Praat, by hand. It will be tedious no matter what. The final analysis of the spectra would have to be done in R or some such, because Praat doesn't have the tools for high level data analysis. The question is about the first part of the data analysis, call it the DSP stage, of playing around with extracting spectra. The question boils down to, are the packages available for R (or other programs) as comprehensive and easy to use as the ones built into Praat? Nov 23, 2019 at 21:45
-
You can also make this kind of analysis in Praat. Don't underestimate Praat. If you don't have many items, you can even make it manually, otherwise you have to learn Praat language, which is not easy. Coding is only interesting if your data is large. Yes they are easy, examples: medium.com/@taposhdr/… Nov 23, 2019 at 22:50