Transcript
Page 1: Computational Neuroscience Lecture 7

Computational Neuroscience Lecture 7

Conor Houghton

Page 2: Computational Neuroscience Lecture 7

PICTURE FROM WIKIPEDIA

Page 3: Computational Neuroscience Lecture 7

The inner earPICTURES FROM WIKIPEDIA

COCHLEA

CROSS SECTIONOF THE COCHLEA

Page 4: Computational Neuroscience Lecture 7

Outer hair cells amplify

http://youtube.com/ /watch?v=Xo9bwQuYrRo

• Video of dancing haircell.

Page 5: Computational Neuroscience Lecture 7

Inner hair cells signal

http://youtube.com/ /watch?v= 1VmwHiRTdVc

• Video about the inner ear, with the sound removed.

Page 6: Computational Neuroscience Lecture 7

Stereocilia of a frog’s inner ear

PICTURE FROM WIKIPEDIA

Page 7: Computational Neuroscience Lecture 7

Different hair cells respond to different frequencies – all hair cells respond to sound over a short time window.

Page 8: Computational Neuroscience Lecture 7

This gives a windowed Fourier transform.

Page 9: Computational Neuroscience Lecture 7

Windowed Fourier transform

Page 10: Computational Neuroscience Lecture 7

s(t)

Page 11: Computational Neuroscience Lecture 7

k(t)

Page 12: Computational Neuroscience Lecture 7

s(t)k(t)

Page 13: Computational Neuroscience Lecture 7

Weber’s law

• Roughly speaking – effect goes like the log of the cause.

• Sort of holds for the auditory system.• Use log|S(k,t)|

• SMALL PRINT: The phase information is gone, however, we have overlapping windows and two variables; there are theorems that say we haven’t lost anything.

Page 14: Computational Neuroscience Lecture 7

Spectrogram

http://youtube.com/ /watch?v= 5hcKa86WJbg

• Zebra finch song and spectrogram.

Page 15: Computational Neuroscience Lecture 7

Spectrogram

http://youtube.com/ /watch?v= 5hcKa86WJbg

• Repeat of zebra finch song.

Page 16: Computational Neuroscience Lecture 7

Zebra finches

http://effieex3.tumblr.com/post/20369494508

Page 17: Computational Neuroscience Lecture 7

Zebra finch brain

Page 18: Computational Neuroscience Lecture 7
Page 19: Computational Neuroscience Lecture 7

Maybe it’s like vision?

Page 20: Computational Neuroscience Lecture 7

Linear model

Page 21: Computational Neuroscience Lecture 7

Error

Page 22: Computational Neuroscience Lecture 7

Minimize error

Page 23: Computational Neuroscience Lecture 7

Calculate the STRF

Page 24: Computational Neuroscience Lecture 7

From Sen et al.J Neuro 2001

Page 25: Computational Neuroscience Lecture 7

From Sen et al.J Neuro 2001

Page 26: Computational Neuroscience Lecture 7

From Sen et al.J Neuro 2001

Page 27: Computational Neuroscience Lecture 7

So?

• Works better than you might expect, particularly in the lower part of the pathway.

• Does not give the whole story, particularly further up the pathway.

• The calculation is hairy, but seems to work, certainly don’t try to improve it.

• The STRFs aren’t quite as revealing as you’d expect.


Top Related