1.3.

1 Android Tab Overview

Fig 1.8 Samsung Galaxy Tab: An Android Tab

The Android Tab is portable owing to its 7 inch screen size and light weight. Light weight also means that it can be carried anywhere. Easy availability of the Android tabs makes them a favorable choice. The tab has video input resolution of 480p (720 X 480 pixels), maximum acquisition frame rate of 30 fps, 1GHz CPU clock speed, 512 MB RAM.

1.3.2 Implementation

Android SDK implementation using Eclipse IDE:

INSTALLATION OF JDK:

Android and Eclipse both work on Java platform. JDK (Java Development Kit) is thus needed to be installed first for the working of both the android applications and Eclipse. After the JDK package is downloaded, the same is installed for the given system.

INSTALLATION OF ANDROID SDK FOR ECLIPSE IDE:

The android open source provides ADT plug-in for eclipse workspace. It needs to be installed in order to use Eclipse for the development of applications in android. The Install new software option under help option in the Eclipse helps in this aspect. After the installation of the android SDK and the ADT plug-in for the Eclipse IDE the environment is ready for the application development using android.

HIERARCHICAL STRUCTURE:

Development using Eclipse workspace has its own advantages. The Eclipse workspace arranges the various folders and the files in a hierarchical order and thus it becomes easy to locate the files in the workspace.

VARIOUS FILES ASSOCIATED AND THE ONES THAT NEED TO BE EDITED:

The main files that help in building the application are the *.java file which is the main java file located in the SRC folder. The other files that need attention are the main.xml (for the layout of the application) and the AndroidManifest.xml (for the various permissions and the features to be incorporated into the application).

RUNNING THE PROGRAM ON THE VIRTUAL EMULATOR:

The program for eye-detection is based on the pre-defined classes of the Android hierarchy viz. FaceDetector.Face in the package android.media. This class has public methods namely eyesDistance (Returns the distance between the eyes); getMidPoint (Returns the mid-point between the eyes); and pose (Returns the pose of the eyes). These methods help in the eye-detection using android.

INSTALLATION OF THE .apk FILE ON THE ANDROID DEVICE:

On running the program, various files get generated in the workspace. These files are again arranged in a hierarchical order due to the use of the Eclipse IDE. There in the bin folder is a *.apk file which is actually the one of our use. An .apk file extension denotes an Android package (APK) file. This file format, a variant of the JAR format ( Java Archive format that aggregates many files into one and generally used to distribute Java applications or libraries in the form of Java class files and associated metadata and resources), is used for distributing and installing bundled components onto the Android operating system. This .apk file is then transferred to the android tab and then installed on the device itself. Thus the program gets installed and is run to detect the eyes, calculate Perclos and assess the alertness level based on a threshold.

1.3.3 Problems faced

While the pre-built classes for eye-detection are used in android, the pre-built classes have their own difficulties. Though we use them for the better functionality of the program as a whole, the classes themselves take some time for their processing and as a result slow down the whole process. The possible solution to that can be such that various optimization procedures can be applied to the program in order to optimize the whole program and in a way optimizing the class runtime itself. Thus the program would get close to real-time (though being compiled just-in-time) and give a real-time approximate of the whole. One of the more prominent difficulties while dealing with the image processing in Android devices is that the Android devices do not come with High End Cameras. The cameras that are mounted on the Android devices have low frames per second (fps) and as a result the resultant image processed is of low quality and cannot become real time. The possible solution to this problem is to hardware interface high end cameras having higher frames per second with the android devices. Thus the application would have higher frames to process per second and hence would give a more accurate result than when lower frames per second camera is used.

1.3.4 Results

It is found that eyes are getting detected in the tab with a good accuracy rate. Some false alarms are also present which suggests a need of fine tuning of the algorithms. The detection rate is also found to be very slow which also leaves scope of improvement in speed of detection. Figure 1.9 shows an instance when open eyes have got detected in real time.

Fig. 1.9 Open eye detection in Android tab