hfesg / evmhr

欧拉视频放大算法安卓（Android）端APP

EVMHR

Introduction

EVMHR is an Android app for detecting human heart rate by analysing the video of face taken from the front camera of the phone, it is based on the theory of Eulerian Video Magnification raised by Wu, H.-Y., et al in 2012. And the implementation of the core function is provided by Weizhou Pan in his blog article: Eulerian Video Magnification.

Here is a simple demo:

Please hold your phone steadily and make your face fill the screen, and try to use this app at a place with good lighting condition.

How to use

Android Studio

Android Studio 2.2 is recommended.

This project contains a large amount of C++ code, so we use NDK to build the native library, which will be called by android with JNI(Java Native Interface). Android Studio 2.2 provides a good compiling compatibility with C++ code, but we are not sure about the success of compilation in previous version of Android Studio.

FFmpeg

The video recorded from a phone camera was compressed by the system, while OpenCV cannot read the video codec used by Android, which will typically be avc or h.264. So we use a compiled version of FFmpeg: ffmpeg-android-java to convert the video to MJPEG format.

To use this library, make sure to have the following line in your app's build.gradle:

dependencies {
	...
	compile 'com.writingminds:FFmpegAndroid:0.3.2'
	...
}

OpenCV

This project is based on OpenCV4Android 3.1.0, you can download it from here. We only need the native folder, so copy it into \src\main\cpp.

NDK

We use NDK 12, just download it from the manager in Android Studio.

Add the following lines to your app's build.gradle:

externalNativeBuild {
	ndkBuild {
		path file("src\\main\\cpp\\Android.mk")
	}
}

If you add your own .cpp files, don't forget to add them into LOCAL_SRC_FILES

If you want to change the native interface function in C++, make sure you follow the naming rule:

Java_ + PackageName + ClassName + InterfaceName

and use InterfaceName in your java code.

License

MIT License