Listening Flashcards
where do the initial stages of acoustic analysis happen?
superior temporal cortex
where does speech processing split into two streams
after inital cortical stages of acoustic analysis speech processing splits into ventral (speech recognition) and dorsal (speech perception) pathways
what is speech recognition
temporal regions of the brain map sound onto meaning (ventral processing stream)
lexical
what is speech perception
parietal/frontal regions map sound onto action (dorsal processing stream) phonological
what does the dual processing stream model predict
that damage to the ventral stream will impair comprehension while damage to the dorsal stream will impair processing repetition of speech sounds
dissociation between performance on auditory comprehension tasks and sub-lexical perception tasks
e.g unable to distinguish cat and cot but can appropriately match these to pictures
parallel pathways with some crosstalk
what is the segmentation problem
how are we able to perceive speech when there are few gaps in words visible in a waveform of a speech signal
e.g perceived gaps are difficult to identify within a waveform or spectrogram as there are no clear boundaries in a continuous stream yet we are able to break it down into words, syllables and phonetic features
thus there is no direct correspondence between auditory energy and perception (no correlation of amplitude and gaps)
how can we identify vowels in a spectrogram
the relative spacing of steady state portions of a spectrogram
how do we perceive speech
the continuous speech signal is perceived as segments
25-50 phonetic segments per second (4-5 syllables)
speech sounds should be identifiable by different combinations of formant frequencies and the shapes of these transitions
what information is there in a spectrogram
the frequencies of sounds produced over time
formants: bands of acoustic energy
steady state: vowel identity
transitions: consonant identity
what is meant by ‘steady state’
the steady state ratio of the first and second formant distinguishes vowels
what is meant by formant transitions
correspond to production of constant
second formant anticipates the frequency of the vowel that will be produced next
what is coarticulation
the features that characterise speech sounds occur non-linearly
we produce them in parallel rather than in sequence
anticipating what next production will be is reflected in production of a sound
overlap of movements in phoneme production
makes speech production fast and efficient, but means spectrogram will be different depending on the context of the following vowel etc
what are the problems for speech perception
non-linearity
lack of invariance
speaker variability
what is meant by non-linearity
acoustic features spread themselves across each other and are not procured as a linear sequence
what is meant by the lack of invariance
speech sounds have more than one waveform representation
the differences between phonemes are very small
there are big differences between different instances of the same phoneme
coarticulation exists whereby the articulation of one phoneme influences that of another
how are the problems of speech perception solved
by perceiving a range of sounds categorically
what is categorical perception
despite the lack of invariance in the acoustic signal we still perceive specific categories that correspond to phonemes (the smallest units of sound which change the meanings of words)
speech perception is categorical as the variation within phonemes is contained in a single category
given categories can’t vary infinitely, what tasks have psychologists designed to better understand the nature of categorical perception?
Voice onset time e.g ga released same time as vocal chords start vibrating whereas ka has a much later voice onset time despite benign produced in the same place
Wig vs wick
Artificial speech syllables
how is voice onset time a continuum
changing the time of consonant release alters the percept
this can happen for both the sound at the start or end e.g ba - ka and wig-wick
what does a place of articulation continuum show
difference between ba and da is place of articulation
this is represented in the slope of the formant (usually second)
changing this incrementally produces a continuum
what happens in categorical perception of speech
hallmarks: identification and discrimination tasks
identification tasks e.g on ba-da continuum produce steep sigmoidal slope characteristic of categorical perception
discrimination task: boundary established with identification task and take stimulus pairs (either same or different side of boundary: cross or within category pairs) - same acoustic unit range each time, asked if they are the same or different - looking for a peak in the discrimination function to demonstrate categorical perception (note this isn’t seen for non-speech sounds, we’re better at discriminating notes on a piano etc)
speech is special
