That completes our main modules on articulatory and acoustic phonetics. You should now have a basic understanding of how vowels and consonants are produced in terms of the vocal tract and it’s articulators. You should also have seen that we can “see” evidence of these articulations in speech acoustics, as represented in a spectrogram. These acoustic cues can be used to “read” spectrograms, i.e. to be able to tell what someone has said by just looking at a spectrogram. This is in essence what automated transcription systems attempt to do! So, it’s important to know what acoustic properties of the speech waveform are important for identifying what has been said for speech recognition. For speech generation, we want to make sure we generate the right acoustic features so that the waveform is understood as speech.
The next two modules will look at the aspects of acoustic phonetics from more of an engineering point of view. We’ll come back to more phon issues in later weeks, as learn more about TTS and ASR. In particular, we’ll look at the source-filter model from both theoretical and engineering points of view.
Making connections between the phonetics material and the speech technologies we’ll look at in the coming weeks will help you be an active learner. Just now, you probably have an understanding of issues in phonetics that will feed into how we design speech technologies, but only a vague idea of the ‘big picture’: the ideas may not yet be well-organised in your mind. Keep connecting and organising, and you’ll find that it does all join together.
What you should know from Module 2
Note: we’ll continue to discuss a lot of the ideas around the frequency domain, resonance and the source filter model in modules 3 and 4.
What does a speech waveform (i.e. in the time-domain) represent?
- Time versus amplitude graphs
- Oscillation cycle
- Period T and wavelength λ (we’ll revisit this in the next few modules)
- Frequency (F=1/T)
- What are “Hertz”?
- How to calculate the frequency of a waveform by measuring pitch periods (Example in the ”waveform” video)
Types of waveform:
- Simple versus complex waves
- Periodic versus aperiodic waves
- Continuous versus transient waves
- Fundamental frequency (F0)
Spectrum:
- The spectrum as a representation of waveform frequency components
- What is the spectral envelope?
- Why do we consider F0 and harmonics to be “source” characteristics
- What the relationship between formants and resonance
- F0 is not a formant!
Spectrogram:
- What do the x and y axes of a spectrogram represent (e.g. in Praat)?
Acoustics of Vowels:
- What is the general relationship between formants (acoustics) to tongue position (articulation):
- F1 and vowel height
- F2 and vowel frontness
- Acoustic vowel space:
- You don’t need to to know the specific formants associated with different vowels, but if you understand the relationship between height/frontness and formants you should be able to deduce this!
Acoustics of Consonants:
- What does voicing look like on a spectrogram or spectrum?
- What are sonorants versus obstruents?
- Identify basic acoustic characterstics (i.e. on a spectrogram) of consonant manners: plosives (i.e., stops), aspiration (for plosives), fricatives, nasals, approximants.
- Clues for place of articulation: stops, fricatives
Voice Onset Time:
- What types of voice onset time are there?
- What does voice onset time look like on a spectrogram?
- How would you measure VOT (i.e. relative to a stop burst)?
Vowel space
- How can you interpret an F1 vs F2 plot of vowel measurements
- Relate this to vowel characteristics/the IPA vowel chart (e.g. tongue height and frontness)
Key Terms
- waveform
- amplitude
- sine wave
- period
- frequency
- wavelength
- Hertz
- fundamental frequency
- harmonics
- spectrum
- spectrogram
- spectral envelope
- Fourier transform
- formant
- vowel height
- vowel frontness
- open vowel
- closed vowel
- sonorant
- obstruent
- plosive
- voicing
- voice onset time
- aspiration
-
- Forum
- Topics
- Last Post
-
-
Acoustics
Sound, how it is produced and how it behaves
- 19
- 6 months, 2 weeks ago
-
Phonetics and speech science
How speech is produced and perceived
- 19
- 1 year, 6 months ago
-
Signal processing
Questions about feature extraction, time and pitch modification, or anything else we can do to speech waveforms.
- 44
- 6 months, 1 week ago
-
Acoustics
