31 terms

12-5-12 speech segmentation

speech segmentation
-breaking connected speech down
new method
Head turn preference proceduring
Infant test as to how infants breakdown phrases
prosodic patterns
-stress patterns in words
-partner vs. depart
-6 months old don't show a preference
-9 month old fro english speaking homes prefered to listen to lists of trochiac words
trochiac words
-words with stress on the first syllable
-rules.probability's governing how phonemes can be combined to form words
sequential phonotactis
-the rules about how phonemes can be combined
probabilistic phonotactics
-which phonemes are common and which phonomes can occur
phonotactic results
-results match those for prosodic patterns
-6 months old show no preference
-9 month olds listen longer to words that
-contain legal combinations
-contain frequent phonemes and combinations
allophonic variation
-the way a phoneme is produced depends on its position in the word and in the surrounding phoneme
-familiarization and testing
-infants listened to passages they heard during the familirization ;phase
-listening to what's appropriate during beginning and end of the word.
phonotactic bootstrapping
-using lagnuage knowledge from one level of language to figure out another level
semantic bootstrapping
-using knowledge or word meangings to figure out grammar
syntactic bootstrapping
-grammatical knowledge to help figure out word meaning
-children have a lot of speech and the beginning and ends of sentences
statistical cues
-listeners can keep track of what units consistently appear together
-8 month old heard a continuous speech stream
-it contained no overt cues to word boundaries
continious stream of utters with no breaks, pitch change, no cues
-infants listened to real words vs. partial words
speech segment
-listeners must learn wher eboundaries are not- it is not innate
speech segmentation
-prosodic, phonotactic, allophonic, statistical cues
-used to figure out what word boundaries are
vowel perception
-measuring speech perception
-stimulus, task, measure, population
speech perception stimulus
-natural speech-can other speakers in other languages understand language.
modified natural speech
how does vowel duration affect perception
synthetic speech
-controlling the frequencies of hte formant
speech perception tasks
-identification- listen to the token
-are the tokens the same or different
-matching- 3 tokens which all the same
3 measure
-response choice-look at perceivers actual responses
-response time- look at how long it takes to perceive to respond
-neural response- EEG to see which brain areas are active during speech perception
clinical-hearing loss, foreign
-developmental status-children adults, elderly
-target language-what languages we use during testing
-listeners language-what is listeners native language
How do we perceive speech
-traveling waves on the choclea are activated at frequencies-dependent locations
basilar membrane
-when presented with speech, the basilar membrane will see peaks that correspond with the formant frequencies
-peaks in energy
-formant transitions
-rapid changes in formant frequencies as articulate
Because of co-articulation
-there are two problems because of co articulation
first problem of co articulation
-unlike letters on a page, there are no pauses between adjacent phoneme
the second problem with co articulation
-tis leads to two related problems
-lack of sementability
-lack of invariance
lack of segmentability
-we cannot break speech streams into individual phonemes
-any segment has info about itself but also about surrounding sounds
lack of invariance
-no neccessary and sufficient attribute uniquely specifies a given speech sound
-beause of these 2 things there are huge problems for acoustic information
-the same acoustic information can signal different sounds