16_Adaptive Noise Reduction Tech

Question 1

Modern HAs use different ______ to deal with different types of noise

Answer 1

algorithms

Question 2

How is internal noise reduced?

Answer 2

expansion

Question 3

What do adaptive noise reduction (ANR) techniques attempt to do?

Answer 3

Separate desired signal (such as speech) from the undesireable noise sources

Question 4

Name 3 reasons we use ANR algorithms when we already have directional mics to reduce noise

Answer 4

• desired and undesired sound sources are spatially co-located
• custom instruments may not accommodate multiple mics
• ANR may complement directional mic processing

Question 5

How is noise managed through gain reduction?

Answer 5

Poor SNR in the low frequencies may mask (through upwards spread of masking) better SNR in high frequencies. By lowering the gain of all inputs in the low frequencies, we may be able to make better SNR frequencies audible higher up

Question 6

How is noise managed through multichannel gain reduction?

Answer 6

• gain in each channel is determined using estimated SNR in that channel
• speech/non-speech detector is used to determine the relative levels of speech and noise in a given channel
• each channel’s SNR is unchanged, but the OVERALL SNR is improved

Question 7

Name 2 ways to separate speech and non-speech

Answer 7

Evaluate the envelope (speech tends to have more variation b/w peaks and valleys than stationary noise sources)

Environmental classifiers (subband energy levels, spectral dynamics, voicing features, synchronicity…)

Question 8

What 2 features can we extract from the envelope that will allow us to detect speech in noise?

Answer 8

Modulation frequency (Hz) 
- dominant frequency range for typical speech envelope is 0.5 - 20 Hz

Modulation depth (dB)
- typical quiet speech modulation depth is 30 dB

Question 9

What environmental classifiers can hearing aids use when modulation features alone are not enough to discriminate b/w wanted and unwatned sounds?

Answer 9

• subband energy levels and their distribution
• spectral dynamics
• voicing features (periodicity, harmonics)
• synchronicity
• etc.
  (then apply gain reduction appropriately)

Question 10

Not all ANR algorithms are created equal. Name 3 possible differences in their implementation

Answer 10

• number of channels
• speech/non-speech decision process
• gain reduction rules
• time constants for activation and release
• interaction c/ other signal processing schemes in the instrument

Question 11

Is there any evidence that ANR algorithms alone either degrade or improve speech intelligibility?

Answer 11

No improvement, and no degradation unless the algorithm is aggressive

• *ANR algorithms do enhance sound quality, and reduce noise annoyance and listening effort, though.
• > better use of cognitive resources for other tasks

May complement directional processing

Question 12

Are ANR algorithms equally effective in reverberant environments?

Answer 12

No

Question 13

How can you verify ANR?

Answer 13

See if the ANR reduces audibility in quiet using speech map

• fitting at 75 dB SPL
• turn on ANR and run again
• if the 2 curves overlap, ANR does not affect speech processing in quiet

Question 14

Which showed a significant improvement in SRT: when NR was added to directional mic on an open fitting or closed fitting?

Answer 14

Closed fitting

Question 15

Do directional mics show significant improvement to SRT compared to omnidirectional mics when used in open fittings, closed fittings, neither, or both?

Answer 15

Both