A.
Results of Crying Preprocess
Fig. 6. shows the results of each step. First, we extract the
raw data. Then we get the data after framing. Next, we
perform the endpoint detection task. The red line indicates the
beginning of the voiced signals and the green line indicates the
end of the voice. Afterwards, we get the data based on the
vocal fragment after cry unit detection. Each data between the
red and green line is a baby crying fragment. And finally, we
get the baby crying signal after cutting and splicing.
B.
Results of Classification
We separated crying data into three types according to the
reason of crying, including hungry, pain, and sleepy. The
number of the preprocessed data is shown in TABLE I.
TABLE I.
N
UMBER OF
D
ATA
S
AMPLE
Kind of Data
Sample
Training Data
Number
Testing Data
Number
Total
Hungry
54
54
108
Pain
47
46
93
Sleepy
47
48
95
Non-Crying
150
150
300
The results of the classifications, which include judging
whether it is crying, sentiment analysis (judging of the baby is
hungry, in pain or sleepy) and a comprehensive analysis in
which we divide the data into hungry, in pain, sleepy and non-
crying is shown in TABLE II.
TABLE II.
R
ESULTS OF
C
LASSIFICATIONS
Dostları ilə paylaş: