Name : Dyah Widhyanti
NIM : 16030214012
Study
Program : S-1 Mathematics
Department :
Mathematics
Faculty : Mathematics and Science
Institution
: Universitas Negeri Surabaya
Advisor
: Prof. Dr. Dwi
Juniati, M.Si.
The sound of crying babies
is one way for babies to express the physical and psychological conditions that
are being experienced. From the sound of crying babies, parents or caregivers
are expected to understand the purpose of the crying. So, research needs to be
done about the characteristics of the baby's crying sound in terms of
mathematical that is based on fractal dimension values. The purpose of this
study was to be able to classify the sound of crying babies based on fractal
dimension values of several infant
conditions using the Higuchi method. The data used in this study amounted to 80
data consisting of 4 types of crying babies namely in conditions of hunger,
fatigue, stomach ache, and discomfort obtained from the website uploaded by Gabor Veres in the GitHub.
The first process of this research is signal pre-processing which consists of
two stages, namely the filtering process followed by the normalization process.
Furthermore, feature extraction using the Discrete Wavelet Transform (DWT)
method with the Daubechies mother wavelet that Db4 with 5 level wavelet
decomposition. From this process, an approximation signal with the time domain
is converted into the frequency domain using the Fast Fourier Transform (FFT)
algorithm. Signal approximation results from the FFT process are then
calculated fractal dimension values using
the Higuchi method. In this study k-max values of
10, 16, and 50 were selected as experiments. The results of the fractal
dimension value of each sound signal are then carried out in the data distribution
process. In this study using the k-fold cross-validation with
. The classification process performed
using the K-Nearest Neighbor (KNN) method and Support Vector Machine (SVM). In this study, several
trials were conducted with
for KNN. Whereas the SVM method uses linear
kernel, polynomial kernel, and RBF kernel then selected value
and as an
experiment. Based on the results of the study obtained the best evaluation that
is equal to 78.75% at the level of decomposition 5 with 5-fold
cross validation, k-max =
10 and the value of K = 9 in the KNN method. The SVM method obtained the best
accuracy of 80% at level 5 decomposition with 10-fold cross
validation, k-max=10 and
and using a RBF kernel with
. So that in this study the SVM method is better than the KNN
method.
Keywords:
Baby cry sound, Higuchi fractal dimension, K-Nearest Neighbor (KNN), Support
Vector Machine (SVM).