Psych 463 Sensation and Perception
Order ID# 45178248544XXTG457 Plagiarism Level: 0-0.5% Writer Classification: PhD competent Style: APA/MLA/Harvard/Chicago Delivery: Minimum 3 Hours Revision: Permitted Sources: 4-6 Course Level: Masters/University College Guarantee Status: 96-99%
Psych 463 Sensation and Perception
Eleventh Assignment (Speech and Sound Localization)
- This question requires that you record speech samples using the program Praatthat you used for the earlier assignment on sound. To record, select “New”, “Record Mono Sound”. You will need either a microphone, a headset with a microphone, or a built-in microphone (which most laptops have). If none of these are feasible, then you can contact me, and I will set you up on a computer in my lab. For all recordings, try to minimize blank time.
When you record sounds, make sure that the 44100 Sampling Frequency is selected
Save each sound recording as a .wav file, and attach the file to the assignment in Canvas.
Once you have your sound, don’t forget to Click “Edit” to get the spectrogram. Note that the spectrogram is the graph in the bottom half of the window, and this is the one you should be focusing on. If your spectrogram has red curves marking the formants, like the following, you can click the Formant pull down menu to toggle “Show Formants”” off.
Select “Spectrogram Settings” from the Spectrum pull-down menu, and set the upper limit to 5000 as seen below:
Copy (Alt-PrintScreen or Fn- Alt-PrintScreen for some laptops) and paste the spectrogram into the document. You can then use the drawing features if you like to point an arrow to the places that you are referring to in your answers.
- Record /pa/ /ba/. Select edit to look at the sound spectrogram (lower portion of the graph).
- First highlight (drag cursor with left mouse button down) and cut most of the balnk space in your recording.
- Highlight on the graph the portion of /pa/ that corresponds to the delayed voicing onset
iii. Read off the measure the duration of the delayed voicing onset (duration of highlighted area). In my example below, the highlighted portion of the delayed voicing onset for /pa/ is .043 seconds or 43 ms. You may have to zoom in on the highlighted area for the number of seconds to show.
- Record boe – doe (as in “bow-tie” and “doe, a deer”, or other sound combinations that differ only on place of articulation, like dee – bee or pea – key). The steady-state part of the formant should be identical. Mark and describe the beginning of the second formant for each syllable.
- Record dee – day- deh (as in deck) – doo – doe – da (as in dad) – and /d^/ (as in duck). Copy and paste your graphs, as I did below for /dee/. Look particularly at the beginning—i.e. the /d/ part. The point is that there is no consistent acoustic “thumbprint” for /d/–i.e. nothing at the beginning of each syllable is consistent for all /d’s/. For simplicity, though, focus on the beginning of the second formant for each syllable, and complete the following table, putting a check mark in the appropriate column for each syllable:
I have recorded “dee” and done the first row as an example. The arrow indicates the beginning of the second formant. Note that your voice may well give somewhat different results.
Starting f of 2nd formant Sharply falls Falls level rises Sharply rises dee 1500 Hz day deh doo doe da d^
- Record shhh – ssss. Try to equalize the intensities, probably by attempting to say ssss more loudly. You will need to change the high frequency range to see the difference. Select the “Spectrum” pull, down menu, then “Spectrogram Settings, then set the high value of the range to 20,000 Hz. Attach the spectrogram at the end.. What distinguishes the two sounds? (HINT: Look at the high-frequency energy at each word). Note that in speech, unlike sounds in general, small differences in the physical energy can sound very different.
- Do the following sound localization demo. You will need two accomplices. Three people should stand in a line, with the middle person about 5 to 10 feet away from the other two. The middle person (preferably you) should be facing one of the other two people, so that the sound will either be coming from the front or from the back, and the other two should be facing each other. This works better outside (to avoid sound reflections off of walls), but can be done inside if weather does not permit. The person in the middle should have their eyes closed. One of the two outside people should be in charge of indicating which of the two will make a sound (like a clap or finger snap, etc.) Vary the sounds and the person making them randomly. See if the person in the middle can tell who made the sound. Repeat ten times. Probably you can do this successfully most of the time, but occasionally you will make a mistake. However, this may not be the case–report on what happens. Now repeat the procedure with the person in the middle holding his/her hands firmly over his/her ears. They should still be able to hear the sound. Report on how many errors are made now. Account for your results with reference to the Head-related Transfer Function (HRTF).
Ears Uncovered Ears Covered Number Right Number Wrong
Discussion of Results:
- (Optional Extra Credit). If you have a video camera and some video editing skills, try to create a video of someone saying /fa/, but dub in /ba/ instead. Report on what you hear (or better yet someone else hears) when watching the video vs. just listening. Please attach a short digital file of the video.