Across Channel Processing Flashcards
1
Q
Across Channel Processing
A
- recall: frequency resolution: the ability to separate out one component in a complex sound, auditory filters
- as auditory filter gets wider it lets in more frequencies around it, signal increases to a point and then equalizes and stays stable past a point, you can isolate and use the information, separate but can talk to each other
- across-channel processing: interactions among or comparisons across separate frequency channels that affect auditory performance
- across channel = across filter = across frequency
2
Q
Examples of across-channel processing:
A
That aid detection: - psychophysical two tone suppression - comodulation masking release (CMR) - profile analysis That impair detection: - modulation detection interference (MDI) - Psychophysical overshoot
3
Q
Psychophysical two-tone suppression
A
- mach bands, making it easier to distinguish the edges
- aka two tone unmasking
- masker tone frequency and level are fixed
- suppressor tone level is fixed, but its frequency varied across tests to form the curve
- The detection threshold for the forward-masked signal is determined:
- in the presence of only the masker tone (horizontal straight line)
- in the presence of the masker + suppressor (points on the curve - look at diagram)
- a lower threshold in 2 than 1 shows that there was a suppression
- threshold for forward masking, masker then signal
- no extra masking from the ‘suppressor’ tone, they are both together but far enough apart that nothing happens, outside of critical band
- pull them in together and there is extra masking, together within critical band
- *low masks high better than high masks low
- “release from masking” when you start to pull it away (under the curve)
4
Q
Psychophysical two-tone suppression: actual stimulus vs internal representation of that stimulus
A
- internally the lower-level masker does less masking
- “big tall guy, pushes down little guy”
- entirely internal, suppresses the neighbor, he isn’t getting taller but the small is getting small so it seems bigger
5
Q
Simultaneous vs forward masking
A
- no two tone suppression in simultaneous masking - threshold only gets worse, or increases frequency still varied the suppressor, suppressing both, no difference between signal and masker
- happens all the time but how do you reveal and show that it is happening
6
Q
Psychophysical two-tone suppression: simultaneous masking
A
- suppressor suppresses both the masker and the signal, no change in signal-to-noise ratio
7
Q
Suppression and hearing loss
A
- measure suppression with narrow vs wide band noise
- more hearing loss, more is lost, you have a harder time because this is supposed to mark the edges, shows you where one filter starts and another stops - exaggerating the differences so you can see the edges, with hearing loss the “mach bands” are more smeared together
8
Q
Comodulation masking release
A
- detect a tonal signal in the broadband masker - signal inside noise in one condition and during the other the noise is modulated
- BW increases, the threshold gets worse but eventually this does drop thresholds - more noise you have the easier it becomes to detect the stimulus
- when it’s comodulated you may pay more attention to the pits/valleys
- Cue band and target band have different frequencies but the envelope is the same, cue band is outside the CBW
9
Q
Release from masking
A
- when threshold goes down
10
Q
CMR and Hearing Loss
A
- comodulation produces independent information, amplitude modulation is within bandwidth
= if you have hearing loss you will loose CMR, bigger is better, you get more release as the bandwidth gets bigger
Why? - already higher levels
- suppression gets worse with hearing loss
- CBW gets wider with hearing loss
- loose independent information when the bands overlap
11
Q
Profile Analysis
A
- centrally mediated aspects
- intensity discrimination: most often for pure tone, has some sort of shape
- x: frequency y: level
- where is the different frequency information? compare the levels, instead you are comparing 2 stimulus levels inside same observation, you are looking at profile or the whole shape level based on the neighbor levels
- comparisons within
- randomized overall level
as you add more information, you have more examples of what that is so you get better and better, the more components the easier it is to detect - signal to detect is always in the center
12
Q
Modulation Detection Interference - Impairs detection of across-channel processing
A
- how close to 0 in depth can you get and still tell there are changes in depth
- listening at probe frequency, not modulated
- detecting modulated tone, example of selectivity because they don’t effect
- now: amplitude modulate then the same as the bands you are listening to - aweful, you have a really hard time,
- not the same modulation - threshold drops again
- same rate vs. different rate: brain is coding by rate, we must also have modulation filters
13
Q
Psychophysical Overshoot
A
- the detection threshold for a simultaneously masked signal is determined at different temporal positions in the masking sound
- threshold is highest when the signal is presented at the very beginning of the masker, this is psychophysical overshoot
- you are moving the signal through different points of time of the noise, beginning middle and end of masker.
- masking is not equal, depends on the time
- 195 ms is classic shape of the critical band, if you turn on and wait this long, good - opens the filter wider
- overshoot magnitude: threshold for 1ms delay minus threshold for 195 ms delay
- more noise you put in the worse the threshold gets, so when you are turning on the noise and then immediately turn on the signal, harder to pick out the signal from the masker
Hyp: a bunch of narrow bands working together
14
Q
Overshoot and Hearing Loss
A
- amount of overshoot -> difference between reaction when signal next to each other and far apart
- in HL the magnitude goes down, don’t have the bad performance in the beginning
- as onset of noise increases, threshold increases (gets worse) because of the critical band
- almost everyone shows this but not present in HL
- broadband noise - high frequency phenomenon
