Loading ...
Sorry, an error occurred while loading the content.
 

getting f0 measures at two

Expand Messages
  • hanmengru@gmail.com
    I have a question about how to get F0 measures in this situation: I have marked the boundaries for a utterance on Tier 1 and boundaries for a word within the
    Message 1 of 3 , Nov 20, 2017

      I have a question about how to get F0 measures in this situation:


      I have marked the boundaries for a utterance on Tier 1 and boundaries for a word within the utterance on Tier 2:

      _________________________________

                |           utterance           |   (tier 1)

      _________________________________

                           |word|                     (tier 2)


      To measure the prosodic prominence of the word in this utterance, I want to get the F0 measures (mean F0, min F0 and max F0) of the utterance EXCLUDING the word.


      I know how to get these measures of marked intervals (i.e. 'utterance' and 'word' in the example), but the F0 measures (especially mean F0) of the utterance excluding the word is difficult to get from existing annotation. What I could think of was to get pitch listing of utterances at every 50ms, then extract the 'word' part. However, I'm not sure whether such extraction could be done automatically.


      I'm stuck here, has anybody tried the same thing or does anybody have any ideas? Or the only solution is to add boundaries for the rest of the utterance?


      Many thanks in advance!

    • Daniel Hirst
      You could do this. Let’s assume you have already got the values of utterance_start, utterance_end, word_start and word_end and that the Pitch object is
      Message 2 of 3 , Nov 20, 2017
        You could do this.
        Let’s assume you have already got the values of utterance_start, utterance_end, word_start and word_end and that the Pitch object is selected.

        Then you can do:

        > utterance_mean = Get mean: sentence_start, sentence_end, "Hertz”
        > word_mean = Get mean: word_start, word_end, "Hertz”
        > mean1 = Get mean: sentence_start, word_start, "Hertz”
        > mean2 = Get mean: word_end, sentence_end, "Hertz”
        > part1 = word_start - sentence_start
        > part2 = sentence_end - word_end
        > not_word_mean = (mean1*part1 + mean2*part2)/(part1 + part2)

        In the last line you weight the mean of the parts before and after the word by the relative duration of the parts.

        For the minimum and maximum you just need to take the min and max of the parts and choose the smaller value of min and the larger value of max

        Hope that helps.

        daniel


        > On 20 Nov 2017, at 23:43, hanmengru@... [praat-users] <praat-users-noreply@...> wrote:
        >
        >
        > I have a question about how to get F0 measures in this situation:
        >
        >
        >
        > I have marked the boundaries for a utterance on Tier 1 and boundaries for a word within the utterance on Tier 2:
        >
        > _________________________________
        >
        > | utterance | (tier 1)
        >
        > _________________________________
        >
        > |word| (tier 2)
        >
        >
        >
        > To measure the prosodic prominence of the word in this utterance, I want to get the F0 measures (mean F0, min F0 and max F0) of the utterance EXCLUDING the word.
        >
        >
        >
        > I know how to get these measures of marked intervals (i.e. 'utterance' and 'word' in the example), but the F0 measures (especially mean F0) of the utterance excluding the word is difficult to get from existing annotation. What I could think of was to get pitch listing of utterances at every 50ms, then extract the 'word' part. However, I'm not sure whether such extraction could be done automatically.
        >
        >
        >
        > I'm stuck here, has anybody tried the same thing or does anybody have any ideas? Or the only solution is to add boundaries for the rest of the utterance?
        >
        >
        >
        > Many thanks in advance!
        >
        >
        >

        ———————————————————————————————————
        alternative emails: <daniel.hirst@...> <daniel.hirst@...>

        CNRS Laboratoire Parole et Langage (UMR 7309) - salle A304
        Aix-Marseille University, 5 avenue Pasteur BP 80975
        tel: +33-4-1355-2712 secr: +33-4-1355-3632/4
        13604 Aix-en-Provence, Cedex 1, France
        ———————————————————————————————————
        Speech Prosody SIG http://sprosig.org/
        Praat-users list: http://groups.yahoo.com/group/praat-users
        My publications: https://www.researchgate.net/profile/Daniel_Hirst
        ———————————————————————————————————
      • Paul Boersma
        ... That is not needed. You can select the desired part by clicking, without there being boundaries on the same tier. It is possible to select stretches
        Message 3 of 3 , Nov 20, 2017
          On 20 Nov 2017, at 23:43, hanmengru@... [praat-users] <praat-users-noreply@...> wrote:

          _________________________________

                    |           utterance           |   (tier 1)

          _________________________________

                               |word|                     (tier 2)



          >  Or the only solution is to add boundaries for the rest of the utterance?

          That is not needed. You can select the desired part by clicking, without there being boundaries on the same tier. It is possible to select stretches between boundaries on different tiers. For instance, you click on the left boundary of "utterance", then shift-click on the left boundary of "word"; the clicking does not have to be very precise; the selection boundaries will automatically snap to the desired interval boundaries if you click in the correct tier. This way you'll get F0 values for the two desired parts exactly. You would still have to decide about how to average the two values.

          Deleting "word" from the sound is not a good idea, because that would make the pitch values near the suture unreliable.
          _____

          Paul Boersma
          Professor of Phonetic Sciences
          University of Amsterdam
          Spuistraat 134, room 632
          1012VB Amsterdam, The Netherlands
          http://www.fon.hum.uva.nl/paul/

        Your message has been successfully submitted and will be delivered to recipients shortly.