Computational Neuroscience 2001b 0368.3013.01
Frequently Asked Questions
1. Should we consider a two ear model or
2. The book didn't explicitly specify any
exact numerical data about the frequency Gaussian parameters of the hair
cells and nerves. Should we assume the parameters by looking at the diagrams,
or is there another source of information?
A. There are, but the diagram should be enough.
3. In the ear the membrane actually transforms
frequency to location (along it), thus activating the correct hair cells.
Is it your intention that we simulate this by extracting the frequencies
right at the start of the process ?
A. Yes, but note that a given hair cell is activated by more than one frequency.
4. Do we need to check faze differences ?
may we assume all frequency start at 0 time?
A. No and No.
5. May we suffice in the simulation that
the output of the first module (representing the hair cells) be a number
frequency of the Spikes emitted, or do we have to output separate spikes (0/1) and the second module should recognize the frequency
A. The FORMER follows the exercise's description, but please read again carefully.
6. All the samples in the cns1.zip were single
channel files. can we assume there will be only one channel?
7. What is the sample rate of the .wav files
given as input?
A. The sample rate is 2^13 = 8192. The files in cns1.zip didn't contain the right sample rate info. It is now fixed. Please reload it.
8. How do I get the
sampling rate of a wav ?
A. Look at: help wavread.
9. You requested up
to 10,000 Hz in range, but at this sample rate it might be a problem wouldn't
it? (we should sample at least twice the tone frequency).
A. Oops. You are quite right. Use a range up to 4,000 Hz.
10. May we overlook
the intensity factor in the simulation (given that we only need to produce
a frequencies list). For example - do we
need to recognize two frequencies if a 200 Hz frequency is overridden by a different intensity 400 Hz frequency ?
A. Do it to better represent your ear.
11. What exactly are
the diagrams you referred to in faq no 2 ? (I have the 3rd edition).
A. You can check this at the 4th edition, available at the social sciences library (at list one copy reserved).
12. Are we suppose
to use built - in matlab functions in order to separate the frequencies
or are we suppose to use our own methods in order to identify the frequencies?
A. Any of the above will do, but justify yourself (see project definition, for both functions you write).
13. Question 1b - what
do you mean by neuron model ? Can you please give an example of such a
A. You have seen some in class. I can't give you an example here.
14. We have some unsure
issues regarding the "tuned neurons" discussed in chapter 6 (page 166 in
the third edition). One way to understand it is that all those neurons
are identical in their response and ONLY THEIR LOCATION in relation to
the basilar membrane causes their different tuning. The second way to understand
it is that differently tuned neurons differ in their PHYSICAL FEATURES
, and this causes their different "tuning".
A. The former, but this fact is less important for your model.
15. We didn't understand
the explanations of sub tones.
A. A good way to start is to recall the (right) sum of two sine functions.
16. Are we suppose
to handle a situation when there is more then one frequency in a given
17. If there are 2
frequencies present with bandwidth frequency difference between them, what
is the smallest bandwidth (in Hz) we should be able to distinguish?
A. If you choose to refer any parameter not explicitly mentioned in the project definition, you may either do it to better represent the human ear, or explain how and from what reasons, your model differs from the auditory system. In any case, explain yourself.
18. For groups of one,
(no tone change), will the length of the .wav file be 1 sec. or can it
A. It might be longer, though this should not be a problem for you. The longest wav will be shorter than 5 sec, for all groups.
19. In the definition
of the addition question (groups of 3) you asked to output the list of
frequencies and for each frequency its start and end time. It is different
from the chords.txt file you have put on your site.
A. You are right. You don't need to output the start and end time of every tone. For exact description look at readme.txt in cns1.zip.
20. Is there an exact
format for the perception.txt file (when to use tabs, when \n etc.) ?
A. No tabs or spaces. Only new line / carriage return (both the pc and the unix versions of new line are accepted).
21. Can we assume that
the matlab environment will be in the project directory during the automatic
22. Should we deal
with two ear features ?
23. Can we assume that
input files will be a multiple of 1 second?
A. No, but all wav files will be of length at least 0.3 sec.
24. (Extra assignment)
Can we assume that a single tone input files will be at least 0.1 sec ?
25. In the exercise
description you have requested 1% precision in the frequency discrimination.
This does not follow psychophysical measures of the accuracy of the ear.
A. The 1% accuracy can be relaxed, in frequencies where human hearing is less accurate than that.
26. Could there be
an error in the file Chords.txt ?
A. Yes. It is fixed now (in cns1.zip).
27. or groups of two
(no tone change) : should our program deal with input file that contains
changing tones (like DoReMe.wav) ?
28. As far as I understood
- there will be only one input file 'stimulus.wav'.
A. You are right.
29. May the matlab
program be consisted of more than one file ?
the example file we have in the zip file be the files used for checking
the exercise ? any other files ?
31. Can I send you
the exercise files instead of putting a copy at tau ?