voting method. Once the model is trained, we use this model to
predict crying states.
We give the pseudocode as follows.
IV.
S
YSTEM
I
MPLEMENTATION
While we already introduced
the system architecture in
Section III, we now provide further details on the components
used in our system.
•
Sensor: The sensors we use include pressure sensors,
temperature sensors, sound sensors and humidity
sensors. The data collected by the pressure sensor does
not need to be processed, and the data obtained by other
sensors all need to be processed to determine the next
action taken by the system. These sensors are all native
sensors of the Raspberry Pi, which means that they are
all compatible with the Raspberry Pi.
•
Raspberry Pi: We use the Raspberry Pi as user
interface to control the hardware. This interface allows
the user to modify some settings, such as the frequency
and
amplitude of crib sway, music played, etc. The
graphical user interface is displayed on an external LCD
display connected to the Raspberry Pi. After connecting
with the pin of the Raspberry Pi expansion board, the
user can set up the device. The GUI is shown in Fig. 5.
The data processing algorithms,
including speech pre-
processing, SVM algorithms, etc, are implemented in
Python language.
•
Server: The servlets on the server are implemented in
Java programming language. There are several servlets
to response particular requests. All servlets are running
on Tomcat which is equipped on an ALiYun server.
Tomcat is a container of servlets that can store all the
servlets and enables them to run on the server. The data
sending to or receiving from servlets are in JSON
format, which is a uniform communication format.
•
Mobile Terminal: So far, we have only developed an
App for Android phones
as Android is the most
commonly used operational system for smartphones.
Java programming language using the Android
Software Development Kit (SDK) has been used for the
development and implementation of the App. The main
functions on this App is showing baby
’s data in graphs
and charts and uploading baby
’s videos as well as
photos. The app also allows parents to set the hardware,
like playing particular music, setting buzzer
’s volume,
etc.
With the help of this App, parents can see the
growth of their baby directly and know the next move
of their baby so that they can make some changes to
meet the baby
’s requirement.
V.
E
XPERIMENTAL
R
ESULTS
The experiment were performed on a computer with an
Intel(R) Core(TM) i7-7700HQ processor and 8GB
of main
memory, running Microsoft Windows 10 Professional.
All crying data we used for the experiment has been
extracted from videos of crying babies that have been shared on
the YouTube platform. Our dataset includes five male and six
female babies of different ethnicity, i.e., three Asian babies, five
Caucasian babies, and three Black babies. Their sentiment is
labeled according to the title of
the video and assessed by a
professional nurse. The non-crying data also comes from the
Internet, including silence, noise, laughter,
chicken roar,
barking, meows, footsteps, etc.
In the remainder of this section we show the results of the
algorithm described in Section Ⅲ, including the result of the
crying preprocess and the result of classification.
对话框标题
Rotation amplitude adjustment
5°
WIFI settings
Music settings
Current connection: none
Current music: none
Volume
Edit
Eidt
Figure 5. User Interface of the Raspberry Pi
Input: f is the feature vector extracted;
C[]
are the candidate classes;
S()
is the trained SVM classifier;
V[]
are the number of votes of the classes;
Dostları ilə paylaş: