CSD 5400 REHABILITATION PROCEDURES FOR THE HARD OF HEARING

1
CSD 5400REHABILITATION PROCEDURES FOR THE HARD
OF HEARING

• Auditory Perception of Speech and the
  Consequences of Hearing Loss

2
Overview

• The aural rehabilitation goal is to remediate the
  effects of a hearing impairment
• Ultimately comes down to the effect of the
  hearing loss on speech recognition and perception
• Develop a general understanding of what a hearing
  loss does to the speech signal

3
The Auditory System in Review

• The primary purpose of the auditory system is to
  take the speech code at the periphery and convert
  it to a representation used by the CNS to extract
  meaning

4
The Auditory System in Review

• Speech arrives to the auditory periphery as a
  series of pressure variations as a function of
  time
• The normal auditory periphery converts these
  pressure variations into physical movement of the
  middle ear structures, which in turn causes fluid
  movement in the cochlea

5
The Auditory System in Review

• Cochlear fluid movement gives rise to the
  traveling wave along the basilar membrane
• Spectral code
• Depending on the site of maximum amplitude of
  displacement of the traveling wave, certain
  auditory nerves will be activated
• Neural activity
• Critical band theory
• Spectral and temporal code

6
The Auditory System in Review

• As the signal moves higher into the central
  pathways, more complicated processing occurs
• Binaural processing
• Temporal processing
• By the time the signal reaches the cortex, it has
  been analyzed and re-coded in a number of
  different ways
• The cortex recognizes these various forms of
  analysis and extracts what is necessary, given
  the job at hand

7
When the Auditory System is Impaired

• Speech is inaccurately coded at the periphery
• Distorted
• Missing
• Attenuated
• Loss of redundancy
• When the signal reaches the cortex, the coded
  representation may be unrecognizable

8
Whos Making Use of the Signal?

• Important consideration
• Adults
• Rely very heavily on the linguistic, contextual,
  and nonverbal cues available
• Children
• No extensive language base

9
Acoustic Cues of Speech

1. Frequency
2. Intensity
3. Temporal Characteristics

10
Flexers Analogy
11
Illustrating Hearing Loss

• Tape examples

12
Acoustic Cues of Speech

• Short Term Characteristics
• Long Term Characteristics

13
Long Term Characteristics of Speech

• Average changes over relatively long periods of
  time
• Provides general acoustic characteristics of
  speech

14
Long Term Characteristics of Speech

• Mean intensity level of conversational speech is
  65-70 dB SPL
• Individual speech segments fluctuate around this
  mean by 40 dB

15
Long Term Speech Spectrum

• Long-interval acoustic spectrum of male voices
  taken 17 inches from speakers lips
• Maximum energy is at approximately 500 Hz
• Roll-off rate of 9 dB/octave

16
Phonemes

• Smallest unit of speech to have linguistic
  meaning
• Traditional unit of speech to study short term
  acoustic characteristics

17
Phonemes

• Classification system
• Vowels
• Consonants

18
Differences Between Vowels and Consonants

• These two classes of sounds differ in the manner
  they are produced and in the way we perceive them
• Vowels are considered more prime
• Rhyming
• Speech Errors
• Vocal tract configuration
• Voicing

19
Short Term Acoustic Characteristics of Vowels

1. Vowels are always voiced
2. The vocal tract is relatively open
3. Source-Filter Theory of vowel production

20
Sound Source of Vowels

• The glottal pulse
• The lowest component is the fundamental frequency
  (f0)
• Harmonics are labeled Hx.
• Maximum energy is at the fundamental frequency of
  the speaker
• Above the fundamental frequency, the spectrum
  rolls off 10-12 dB/octave

21
Filter of Vowels

• The vocal tract, which can be thought of as a
  tube open at one end, closed at the other, and of
  a specified length

22
Putting the Source and Filter Together
23
Putting the Source and Filter Together

• The panel at the left shows the glottal source.
  The panel at the right shows the spectrum of the
  source after filtering by a filter representing a
  neutral vocal tract. The spectral
  characteristics of the filter is indicated in the
  middle panel

24
Changing the Effects of the Filter

• In order to produce these three different vowels,
  we change the characteristics of the vocal tract.
  This will alter the resonant frequency
  characteristics of the tube and change the
  combined spectrum of the glottal pulse and the
  vocal tract

25
Changing the Effects of the Source

• This is what happens when the same vowel is
  produced by a man, a woman and a child

26
An Important Short Term Acoustic Characteristic
of Vowels

• Formants are the regions of increased spectral
  energy
• They are only a characteristic of vowels
• The frequency regions they occupy, as well as
  their relative intensities change as the vocal
  tract changes with each vowel production
• All English vowels have 5-7 formants
• Vowels can be distinguished from one another
  using the lowest (frequency) 2-3

27
Vocal Tract Shapes and Spectra

• Vocal tract shapes and corresponding spectra (F1
  and F2 only) for four back vowels

28
Vocal Tract Shapes and Spectra

• Vocal tract shapes and corresponding spectra (F1
  and F2 only) for four front vowels

29
Peterson Barney (1954)

• Landmark spectrographic study of 76 men, women,
  and children producing vowels in isolation
• Measured and reported the average fundamental
  frequency and the frequency/intensity of the
  first three formants of the ten English vowels

30
A Summary of Peterson Barneys Results
31
Articulation and the Formant Frequencies

• F1 corresponds to the degree of tongue
  constriction in the vocal tract
• F2 corresponds to how forward in the mouth the
  tongue is
• F3 is not related in a simple way to articulatory
  parameters

32
Vowel Normalization

• Vowel quadrilaterals for men, women, and children
• Whats thought to be important for vowel
  perception is the relative spacing between F1 and
  F2 not their absolute frequencies

33
Consequences of Hearing Loss on Vowel Perception

• Vowel perception is impaired when a hearing loss
  erodes the acoustic information in the F2 range
• Generally 1000 Hz and above
• Vowels are generally robust to the effects of
  hearing loss

34
Short Term Acoustic Characteristics of Consonants

• Differences Between Vowels and Consonants
• Consonants
• Have a shorter duration
• Cant be isolated
• Dont have just one noise source
• Arent static
• Identification seems to rely primarily on the
  vowel that precedes or follows
• Have a variety of methods of production and
  places in the vocal tract where they are produced

35
Spectral Regions of Various Speech Sounds

• A common spectral representation of major speech
  sounds
• Related to the threshold of audibility curve

36
Spectral Regions of Various Speech Sounds

• Another example
• Lines A, B, and C represent three different
  configurations and degrees of hearing loss
• What predictions can you make?

37
Spectral Regions of Various Speech Sounds

• Intensity and frequency distribution of speech
  sounds overlaid on an audiogram
• Predictions based on characteristics of the
  hearing loss

38
Predicting the Degree and Type of Phoneme Errors

• These type of charts are used often to help
  predict the effect of a particular degree and
  configuration a hearing loss might have on speech
  understanding
• This works somewhat, but it only looks at the
  influence of a hearing loss in terms of a filter
• Sensorineural hearing loss is more complicated
  than this
• Attenuation and distortion

39
Hearing Loss as a Loss of Redundancy

• Illustrates the reduction of pattern details
  (redundancy)

40
The Consonant Classification System

• Every American English consonant can be
  identified uniquely according to its
• Manner of articulation
• Place of articulation
• Voicing

41
Consonant Feature Classification System

• Classification of the consonants of American
  English according to the articulatory feature
  system

42
Acoustic Properties of Articulatory Features

• Voicing
• Energy is broadband and extends from 100-4000 Hz

43
Acoustic Properties of Articulatory Features

• Place of Articulation
• Energy is very high frequency and confined to
  1000-8000 Hz

44
Acoustic Properties of Articulatory Features

• Manner of Articulation
• Energy is spread through the mid frequencies
  (250-3500 Hz)

45
Consonants and Vowels Together

• Schematic oral tract movements, etc for phrases a
  buy, a pie in the top spectrograms and a dye and
  a tie in the lower spectrograms

46
Formant Transitions

• Schematic of a transition and steady-state
  portion of a formant frequency

47
F2 Formant Transitions

• The second formant transition provides a lot of
  information about the consonant
• Place of articulation is related to the direction
  of the transition
• Manner of articulation is related to the rate of
  the transition

48
Error Patterns with SNHL

• Place of articulation and manner of articulation
  error rates for 38 SNHI listeners
• Place of articulation errors are more prevalent,
  followed by manner of articulation errors

49
Feature Recognition as a Function of Degree of HL

• Auditory identification of temporal patterns of
  vowels and consonant features by 121 HI children
  as a function of PTA
• Notice how place of artic feature recognition is
  adversely affected by HL
• Voicing and vowel id are better preserved

50
Summary of Findings

• General findings of studies of phoneme perception
  for SNHL when using meaningful CVC stimuli
• Relatively few errors are made with the vowel
• When they do occur, they occur more often for
  front vowels
• Higher F2 frequency
• More errors are made with consonants
• Final position is extremely vulnerable
• Most common error type is place of articulation,
  followed by manner of articulation
• Voicing errors are rare

51
In Closing..

• It appears that phoneme error types seem to
  relate somewhat to the frequency region and
  degree of hearing loss
• If the hearing loss is primarily confined to the
  high frequencies, then we tend to see more errors
  with articulatory features that are more high
  frequency weighted (e.g. place)
• Our predictive ability stops here

52
In Closing..

• In fact, we see tremendous variability among hard
  of hearing listeners in terms of their ability to
  perceive and understand speech
• The amount and way information is coded varies
  from listener
• Varying degrees of distortion not related to the
  characteristics of the audiogram
• If information is restricted at the phonetic or
  spectral level, it is also probably restricted at
  the linguistic level
• How well individuals are able to integrate
  information varies

53
(No Transcript)