Transcript Pattern Recognition (cont.) - NAU jan.ucc.nau.edu web server

Pattern Recognition (cont.)
1
Auditory pattern recognition
• Stimuli for audition is alternating patterns of high and low
air pressure called sound waves
• The bigger the difference between high and low pressure
the louder the sound is perceived to be
• The number of cycles per second determines tone
• Humans can hear sounds as low as 20 cycles/sec and as
high as 20,00o cycles/sec
– Other animals have a greater range of hearing, but humans
discriminate between sounds and interpret sound better and
more quickly
– Speakers produce a continuous string of speech sounds that
listeners easily break into separate words
2
Echoic or auditory sensory
memory
• Similar to visual sensory memory in many
ways
• Definition – brief memory system that
receives auditory stimuli and preserves
them for some amount of time
• Since hearing is temporal event, you
would not be able to understand auditory
information if it disappeared immediately
3
Experiments demonstrating echoic
memory
• 3 eared man study
– 3 auditory stimuli (letters or numbers) presented to
the left ear, 3 stimuli to the right ear, and 3 stimuli to
both ears
– Subjects immediatelyasked to report all the stimuli
they could remember hearing
– Reported about 4 of the 9
– Presented with stimuli again and asked either what
were the 1st, 2nd or 3rd stimuli you heard – partial
report
– Subjects more accurate, over 5 recalled, but not as
accurate as Sperling’s measure of visual sensory
memory covered last week though echoic memory
4
Does echoic memory fade?
• Subjects presented with a list of 9
numbers, 2 numbers per second
• Silent vocalization group read the
numbers; a purely visual presentation
• Active vocalization group – read the
numbers out loud as they were presented
• Passive vocalization group- read the
numbers and listened to a tape recording
of someone reading the numbers
5
Results
• Subjects in all groups remembered about 20% of
numbers in the beginning of the list and about 50% in the
middle of the list
• The passive and active vocalization groups remembered
90% of the last item while the nonvocalization group
remembered 50%
• Suggests that the vocalization groups’ recall was
assisted by a still present memory trace in echoic
memory
• Echoic memory traces do persist over time, but as
indicated by decreased memory for early items, the
memory trace fades over extended time
6
Does erasure occur in echoic
memory?
• Backward masking seemed to “erase” meory in visual sensory
memory
• Previous experiment run again. This time after hearing the list,
vocalization groups heard eith a tone or the word “zero”
• Told this was a cue to begin recall, actually used to see if hearing
another word would erase the memory of the last item
• Group that heard the tone performed the same as before. Group
that heard zero decreased accuracy to 50% same as nonvocaliztion
group
• Conclusion: memory of last item was erased.
• Overall conclusion: echoic memory similar to visual sensory
memory, but has a samller capacity and a longer duration
7
Theories of Auditory pattern
recognition
• Largely studied in speech, but some work has
been done with music
• Template theory – we store models or templates
of sounds, an then compare sounds we hear to
sounds we have stored in memory.
• Doesn’t work any better than it did in visual
sensory memory.
• Speech sounds do not always sound the same
though we perceive them as being the same
8
Feature detection theory
• We detect basic features of the sound and use
memory to identify them – much more
successful
• Clearly shows the importance of context
– Phonetic restoration effect – when a speech sound is
missing or covered with white noise, we fill it in and
don’t notice it missing
– Words are more easily recognized when parts are
missing if they are in sentences than as isolated
words
9
Haptic pattern recognition
• Use of somatosensory system
• When blindfolded, most people can
identify objects quickly and accurately
using haptic (active touch) information
• Exploration of an object with hands does
not involve random movements
10
Object recognition using haptic
information
• Begins with enclosing smal object n
fingers and palm
• Then different kinds of stereotypical hand
movements called exploratory procedures
11
Exploratory procedures
• A lateral motion to determine texture.
• Enclosure to determine global shape and
size
• Unsupported holding to determine weight
• Conture following tto determine exact
shape
• Applying pressure to determine hardness
• Static contact to determine temperature
and others
12
Examples
• People reported that shape was most
important in identifying a pencil
• Shape and texture important in identifying
a crayon
• Shape and size important in identifying a
used pencil
13
Visual and haptic information
• Vision not necessary for these exploratory
procedures to develop.
• Visual and haptic objective recognition
processes need to work together
– Using haptic information to identify objects of
different shapes that are the same texture is
slow and difficult
– Visually sorting objects that identical in shape
but differ in temperature or hardness is
difficult if not impossible
14