Presentation

Question 1

Front slide

Answer 1

My name is Gitanjali Sharma and today I will talk about my dissertation project which is called predicting speech-in-noise ability from non-speech stimuli.

Question 2

Outline

Answer 2

This presentation on my project will cover its background, research aims, methods, main findings, brief discussion and conclusion.

Question 3

SiN Perception (3)

Answer 3

• During speech-in-noise (SiN) perception, listeners attempt to identify and understand a speaker’s speech among background noise.
• Most of us have been in this situation before, like listening to a friend in a noisy restaurant.
• In order to perceive SiN , the listener must separate target speech from background noise

Question 4

SiN Difficulty (2)

Answer 4

• It is common for hearing-impaired individuals to have trouble perceiving SiN
• A number of tests to diagnose SiN difficulty has been developed
  *

Question 5

SiN Tests

Answer 5

• One of these tests in SiN tests which plays series of speech stimuli (such as words or sentences) against babble noise and asks participants to report the target speech.
• Try to figure out the sentence in your mind as I play an example of SiN tests?
• Did you guess correctly? Here’s is the sentence:

Question 6

SiN Tests Disadvantage (3)

Answer 6

• The main disadvantage of these SiN tests is that it lacks generalizability to the general population as the speech content is recorded by a native speaker with a specific accent
• There will be an undue performance advantage for participants who speak the same language and accent as the recorded speech content.
• People who don’t might do poorly, resulting in bias in identification of SiN difficulty.

Question 7

SFG Stimulus (4)

Answer 7

• Stochastic Figure Grounds (SFG) were developed to assess SiN perception.
• You can see on the screen that the SFG stimulus is composed of set of tones with fixed frequency over time (the ‘figure’) against background of tones that are randomized in frequency and time.
• It sounds like this:
• As SFGs have no linguistic content, they can be used by individuals from a wide range of backgrounds and overcome the limitations of SiN tests

Question 8

Holmes and Griffiths (2)

Answer 8

• Holmes and Griffiths aimed to validate this SFG stimulus as a SiN predictor
• Participants had to say which of two SFG stimuli has a gap in the figure portion in a same-frequency figure discrimination task they created and completed SiN test.
  *

Question 9

Holmes and Griffiths Finding (2)

Answer 9

• Results showed that participants’ figure-discrimination task performance significantly correlated with SiN test performance, indicating validity of SFG stimulus
• This relationship was significant, but the correlation value was moderate, which suggests SFG can be improved to stimulate speech more closely

Question 10

Speech is Harmonic and Dynamic ! (4)

Answer 10

• Taking a closer look at speech, it has two aspects:
• It has a harmonic structure meaning it contains multiples of fundamental frequency
• For example, if a fundamental frequency Is 200 Hz then second harmonic is 400 Hz, third harmonic is 600 Hz etc…
• In addition, speech is dynamic in that frequencies of speech content change over time and we don’t speak with a monotonous voice.

Question 11

Research Aims (2)

Answer 11

• Figure in SFG stimulus is not dynamic (has fixed frequencies over time) and does not contain harmonic structure which may be needed for it to closer mimic speech better.
• Thus in the study we aimed to create a new version of SFG which figure is dynamic and contains harmoncity, tested its validity in predicting SiN test performance

Question 12

New Stimulus (1)

Answer 12

Our new stimulus, the harmonic roving SFG (HRSFG) has two versions: on left is low-frequency harmonic auditory figure and on right is high-frequency

Question 13

Methodology (2)

Answer 13

• This study was conducted in a sound-proof booth with 54 participants recruited via opportunity sampling.
• The participants completed computer tasks that included sentence and word in noise [WiN] test, two versions of pattern discrimination tasks, and a same-frequency figure discrimination task , that used convential SFG, taken from Holmes and Griffiths study.

Question 14

Pattern Discrimination Task (2)

Answer 14

• A pattern discrimination task involved hearing two examples of the HRSFG stimulus and deciding whether they were the same pattern or different patterns by pressing 1 or 2.
• Separate blocks were ran for the high and low-frequency versions

Question 15

Main Findings I (2)

Answer 15

• We found that how well the participants were able to detect low and high frequency HRSFG figures were significantly positively correlated with how well they can detect sentences amongst background noise
• HRSFG performances at high and not low frequencies significantly increased explained variance in sentence-in-noise performance compared to conventical SFG performance

Question 16

Main Findings II (2)

Answer 16

• Low and high HRSFG performance negatively correlated with WiN performance.
• As compared to SFG performance, high and low frequency HRSFG independently contributed more to variance explained by WiN performance.

Question 17

Discussion and Future Work (5)

Answer 17

• Our roving stimulus was significantly associated with Word and sentence recognition performance in noise, indicating that it is a valid SiN predictor
• As compared to SFG stimulus, HRSFG stimulus explains more variance in SiN tests, indicating that it is a better predictor of SiN performance
  The limitation of the study is that qe only included participants with normal hearing ability by testing them before the main experiment and those who spoke English natively. Since our results aren’t applicable to those with hearing deficits and/or aren’t native English speakers, the roving stimulus must be evaluated for its ability to predict SiN performance in these populations.
• Research shown that SiN tests engage brain areas that do or do not overlap with conventional figure-ground perception.
• Since our stimulus predicts SiN performance better than SFG, this overlap between SiN tests and our new stimuli is likely to increase = future research is needed to confirm this.

Question 18

Conclusion (1)

Answer 18

Overall, the study showed that this HRSFG stimulus is validated against SiN performance and can better predict performance in SiN tests than static SFG stimulus. What brain mechanisms engaged with this new stimulus need to be confirmed by neuroimaging techniques.