Download here - University of California, Riverside

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Cluster analysis wikipedia , lookup

Transcript
Mining Mouse Vocalizations
Jesin Zakaria
Department of Computer Science and Engineering
University of California Riverside
Mouse Vocalizations
100
kHz
laboratory
mice
40
124
Time (second)
125
Figure 1: top) A waveform of a sound sequence produced
by a lab mouse, middle) A spectrogram of the sound,
bottom) An idealized version of the spectrogram
The intution behind symbolizing the spectrogram
Figure 1:
top) Two 0.5 second spectrogram
2
representations of fragments of the vocal output of a male
mouse. bottom) Idealized (by human intervention) versions
of the above
Figure 3: The two fragments of data shown in Figure
2.bottom aligned to produce the maximum overlap. (Best
viewed in color)
C
Q
A
X
X
P
A
X
X
P
Figure 4: The data shown in Figure 2 augmented by labeled
syllables
Background
120
Time (sec)
91.1
original
kHz
90.1
Figure 5: A snippet spectrogram that has seven syllables
0
110
idealized
30
76.3
Time (second)
78
Figure 6: top) Original spectrogram, bottom) Idealized spectrogram
(after thresholding and binarization)
4
1
3
4
8
1
1
7
8
Figure 7: left) A real spectrogram of a mouse vocalization can be
approximated by samples of handwritten Farsi digits (right). Some
Farsi digits were rotated or transposed to enhance the similarity
Extracting syllables from spectrogram
connected components
SP
I
L
Figure 8: from left to right)snippet spectrogram, matrix corresponding to an
idealized spectrogram I, matrix corresponding to the set of connected
components L, mbrs of the candidate syllables
A
B
I
J
C
K
D
L
E
M
F
N
G
O
H
P
Editing Ground truth
Figure 9: Sixteen syllables provided by domain experts
a
b
A
c
B
I
C
J
g
d
K
D
e
E
F
L
M
N
h
i
j
f
G
O
H
P
New
Class
k
Figure 11: Ambiguity reduction of the original set of syllable
classes. Representative examples from the reduced set of
eleven classes are labeled as small letters
Editing Ground truth
Classification Accuracy
for edited ground truth
1
for all the labeled syllables
0.8
0.6
0.4
0.2
0
0
100
200
300
400
500
600
700
Adding more instances
Figure 10: Thick/red curve represents the accuracy of classifying
syllables of edited ground truth. Thin/blue curve represents the
accuracy of classifying 692 labeled syllables using edited ground truth
Data mining Mouse Vocalizations
ddcibfcd
dcibfcd
ciaciaci
ciaciaci
ecccccc
eccccccc
ccccccgc
ccccccgc
Clustering mouse vocalizations
Figure 12: A clustering of eight snippets of mouse vocalization
spectrograms using the string edit distance on the extracted
syllables (spectrograms are rotated 90 degrees for visual clarity)
Figure 13: A clustering of the same eight snippets of mouse
vocalization shown in Figure 12 using the correlation method. The
result appears near random
Data mining Mouse Vocalizations
Similarity search / Query by content
i
i
query image
c
a
ciafqcicia
Edit dist 2
b
q
c
i
a
c
a
ciqbqcaacja
Edit dist 3
Figure 14: top) A query image from [1], The syllable labels have been
added by our algorithm to produce the query ciabqciacia, bottom)
the two best matches found in our dataset; corresponding symbolic
strings are ciafqcicia and ciqbqcaacja, with edit distance 2 and 3,
respectively
query image
c
c
c
c
Figure 15: top) The query image from [2] was transcribed to cccc.
Similar patterns are found in CT (first row) and KO (second row)
mouse vocalizations in our collection
[1] J. M. S. Grimsley, et al., Development of Social Vocalizations in Mice. PLoS ONE 6(3): e17460 (2011).
[2] T. E. Holy, Z. Guo, Ultrasonic songs of male mice, PLoS Biol 3(12): e386, (2005).
Data mining Mouse Vocalizations
motif
194.8 – 195.2 sec
944.7 – 945.2 sec
Figure 16
1: A motif that occurred in two different time
intervals of a vocalization. The left and right one
correspond to the symbolic strings ciaciacia and
(log scale)
# of substrings
ciacjacia
40
30
118
44
20
18
16
11
10
0
Assessing Motif Significance using z-score
3983
0
0.5
c
i
c
i c
a
j
a c
1
a
i
a
1.5
Z-score
i
c
2.5
3
3.5
a
motif 1
ia
c
2
motif 2
c
b
c
c
c
c
c
c
c
c
q
c
g
c
g
c
c
c
Figure 17
1: top) Distribution of z-scores, bottom) two sets of
motifs from spectral space with a z-score of approximately
two and three, respectively
Contrast set mining
Overrepresented in
Knock-out
Overrepresented in
Control
Figure 18: Examples of contrast set phrases. top) Three examples of
a phrase ciacia that is overrepresented in KO, appearing 24
times in KO but never in CT. bottom) Two examples of a phrase
dccccc that appears 39 times in CT and just twice in KO
using information gain