W6 Sound Processing Flashcards

1
Q

Define “Psychophysics”

A

How the perception of a stimulus depends on its physical parameters

Example: Loudness

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Compare the Physical and Perceptual attributes of loudness of acoustic pure tones.

A

Physical attribute:

  • Pressure P in Pa
  • Reference sound pressure PO = 20 uPa
  • Sound Pressure Level (SPL) in dB = 20 log (P / PO)

Perceptual attribute:

  • Loudness L in sones
  • 1 sone = loudness of a 1000 Hz tone at 40 dB SPL
  • Loudness power law: L ∝ P0.6
  • 10 dB increase in SPL => doubling of loudness
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Compare the Physical and Perceptual attributes of loudness of electrical pulse trains.

A

Physical attributes:

  • Current I in microamps
  • Phase width PW in microseconds
  • Pulse rate in Hz

Perceptual attributes:

  • Typically PW and pulse rate constant; current varies
  • Loudness power law: L ∝ I2.7
  • Nucleus CI:
    • Current I is an exponential function of 8-bit current level c
    • Log(loudness) is a linear function of current level
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What are Threshold and Comfort levels?

A
  • Threshold (T): the lowest current that is just audible
  • Comfort (C): the highest current that is comfortable
  • T and C vary between implant recipients and across electrodes
  • Depends on:
    • Number of surviving neurones
    • Distance from electrode to neurones
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Compare voiced and unvoiced speech sounds

A

Voiced

  • all vowels, some consonants (e.g. b, d, m)
  • periodic waveforms: Pitch (F0)
    • intonantion (question/statement)
    • tonal languages (e.g. Mandarin)
    • singing
  • formants (F1, F2): spectral peaks
    • identify vowels
  • most energy at low frequencies

Unvoiced

  • some consonants (e.g. p, t, s, sh)
  • unstructured spectrograms
  • most energy at high frequencies
  • lower amplitude
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Compare Normal and Electrical Hearing in terms of:

  • Spectral resolution
  • Temporal resolution
  • Dynamic range
A

Spectral resolution:

  • Normal
    • Tonotopic - place pitch
    • Vowel recognition
  • Electrical
    • Stimulation site
    • Want many channels

Temporal resolution:

  • Normal
    • _​_Detect transients and changes in amplitude
    • Consonant recognition
  • Electrical
    • _​_Amplitude pattern on each channel
    • Want high pulse rate on each channel

Dynamic range:

  • Normal
    • > 100 dB between loudest bearable sounds and softest detectable sounds
  • Electrical
    • Typically 10 dB between Comofrt level (C) and Threshold level (T)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Describe the 5 components involved in the Signal Path from the Microphone to the Implant

A

(microphone)>>

Front End

Filterbank

Sampling and Selection

Amplitude Mapping

RF Encoder

>>(implant)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Compare Omnidirectional, Directional and Dual microphone types

A

Omnidirectional:

  • single port
  • equally sensitive to all directions

Directional:

  • Two ports
  • Most sensitive to front
  • Inherent pre-emphasis (boosts high frequencies)
  • Used in older processors

Dual:

  • DSP combines the two microphone signals
  • Can vary the directionality
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Describe the components of the Front End stage of the Signal Path

A

Pre-amplifier

  • Microphone: 1 kHz tone at 74 dB SPL gives 3.5 mV peak
  • Provides 20 dB gain

Analog-to-Digital Converter

  • To allow Digital Signal Processing (DSP)
  • Nucleus sound processors: 16 bits at 15 kHz

Automatic Gain Control

  • To cope with different sound levels (e.g. user’s own voice 20 dB louder than friend’s)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Describe the components of the Filterbank stage of the Signal Path

A

Band-Pass Filter (BPF)

  • Passes one band of frequencies
  • Blocks other frequencies

Filterbank is a set of BPFs

  • Divides the sound into N frequency bands
    • CIS: typically 6 wide filters
    • ACE: typically 22 narrow filters
  • One filter allocated to each channel (electrode pair)
  • Frequency spacing:
    • Linear spacing below 1000 Hz
    • Logarithmic spacing above 1000 Hz
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Describe the components of the Sampling and Selection stage of the Signal Path

A
  • Filterbank provides N channels output
  • Implant performs sequential stimulation
  • Processor needs to sample the channels
  • Sound coding strategies:
    • CIS
    • ACE
    • MP3000
    • HiRes
    • FSP
    • HiRes120
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is CIS?

A

Continuous Interleaved Sampling

  • First developed for Ineraid implant (Blake Wilson, 1991)
    • 6 channels, percutaneous plug
  • Samples each filter envelope at fixed rate
  • Pulses interleaved across channels
  • Typical channel pulse rates: 500 - 3500 pps
  • total pulse rate - channel pulse rate * number of channels
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is ACE?

A
  • Based on Spectral Maxima Sound Processor (SMSP) (Hugh McDermott, Uni Melbourne)
  • Selects M largest filter bands (maxima) out of N channels at rate R (peak rate per channel)
    • total rate = M*R
    • CIS has M = N

Two benefits:

  1. Reduces total pulse rate –> saves power
  • e.g. CIS vs ACE:
    • CIS: 1000 pps * 20 channels = 20000 pps
  • -> 50 us pulse period –> 20 us phase width
    • ACE: 1000 pps * 10 maxima from 20 channels = 10000 pps
  • -> 100 us pulse period –> 40 us phase width
    • Double phase width –> half the current –> half the voltage rail
  1. Simple form of noise reduction
    * If SNR is positive, omitting the smallest pulse improves the SNR
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Describe the components of the Amplitude Mapping stage of the Signal Path

A

Each channel has:

  • Channel gain
    • Static: allows the spectral shape to be adjusted
    • Automatic: vary gain based on signal (e.g. ADRO, ClearVoice, SNR-NR)
  • Loudness Growth Function (LGF)
    • Match electrical loudness to acoustic loudness
    • Compresses the filter output
  • T and C mapping
    • Maps all sound between Threshold level (T) and Maxmimum Comfortable level (C)

(see image)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

The ProTouch prosthetic hand (Motion Control Inc) is a neural-enabled hand prosthesis collaboration between Florida International University, Motion Control Inc and Cochlear Ltd.

Discuss the facts:

  • who for?
  • how controlled?
  • limitations?
  • project goals?
A
  • For people with a hand amputation (below elbow)
    • < 100,000 arm amputees in the US
  • Presently controlled by a myoelectric system
    • electrode pads in an elastic sleeve, worn on forearm, sense muscle tension
  • Limitations:
    • only one degree of freedom (open/close)
    • awkward to control
    • cannot feel how hard you’re gripping

Project Goals:

  1. Feel how hard you’re gripping
    • sensor signals control stimulation of sensory (afferent) nerves
  2. Control prosthetic hand by thinking about it
    • measure motor control (efferent) nerve firing using Neural Response Telemetry (NRT)
  3. Integrate with Nucleus implant system! (see image)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly