AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |
Back to Blog
Praat script to label vowels11/7/2023 ![]() Is labelled 'C', and vocalic interval 'V'. Is there an automatic aligner that will align/segment based on consonantal and vocalic intervals, where consecutive consonants or vowels are grouped into one interval (e.g. This is the 'acoustic measure extraction' script I would alter if I was to attempt (2) or (3). if vowel followed by vowel, extract value from entire region) Here's the script I'm currently working with. 1) Find an automatic aligner that does the consonantal & vocalic segmentation 2) Create a Praat script that will use the boundaries and labels from the 'phones tier' to create a new 'intervals tier' with consonantal C-intervals and vocalic V-intervals (might require looking at the pronunciation dictionary the Montreal Forced Aligner uses to create phone labels?) 3) Specifying in the Praat script (for acoustic measure extraction) which consecutive labels on the 'phones tier' to extract from? (e.g. Possible solutions I've thought of (not sure if they're all executable). One caveat to note when making consonantal and vocalic boundaries: When a liquid /l, r, w/ precedes or follows a vowel, I want these segments to be included in the same vocalic interval ( /l, r, w/ + vowel ). I'm looking to automate the segmentation and labelling process for consonantal and vocalic intervals. This manual segmentation becomes very laborious for data-sets with large amounts running speech. ![]() However, for consonantal and vocalic intervals, I had to manually create a separate tier labelled 'intervals' and create boundaries between consonantal and vocalic regions (I referenced the phones tier for this). ![]() I used the Montreal Forced Aligner which provided me with word and phone level alignments, resulting in two tiers (see image below)- 'words tier' (top) and 'phones tier' (bottom). To extract accurate values for these measures, my script must ONLY extract from intervals with voicing, not consonantal intervals. My end goal is have one Praat script use an annotated TextGrid with an tier labelled 'intervals' (containing consonantal & vocalic intervals) to extract various acoustic measures (F0, F1-F5, intensity, jitter, shimmer, Harmonics-to-Noise Ratio) from a corresponding. Where every consonantal interval is labelled 'C', and vocalic interval 'V'. ![]() ![]()
0 Comments
Read More
Leave a Reply. |