czsj01 / pre-thursday-assignment

Android Library for device-gestures detection.

DeviceGestureLib

Android Library for device-gestures detection.

1. Distribution

---

The source code is distributed under GNU LGPL ver. 3 (https://www.gnu.org/licenses/lgpl.html)

2. Concept

---

Library analyses accelerometer data and compares current measures with "model" device-gestures. If requirements of similarity met, it generates detection event.

3. Representation

---

Library source code is represented as Eclipse "Android Library Project". There are 2 ways to use it: attach as a Library to your main Android project (Properties -> Android), or use as external jar. All library classes are bundled into info.augury.devicegesturelib package. Classes you require to utilize library are public.

4. Usage

---

At first, you need to create model of gesture. It's better to create model with real accelerometer data, other than put synthetic constants.

Class to receive accelerometer data record:

class DataReciever implements IGestureRecordReceiver {

	public void onResults(float[] side, float[] front, float[] vert) {
		... //result processing
	}

}

Action to record data:

DataReceiver receiver;
...
long interval = 50 * 1000000; //Interval between measures in nanoseconds (50ms)
int count = 10; //Number of measures  

DeviceGestureLibrary.recordGesture(GetApplicationContext(), interval, count, receiver); 

Now we can create up to 3 Axis, which are components of gesture model. Example for front axis:

float[] frontAxisRecord; 
...
float requiredProximity = 0.75f; //Threshold of detection
CompareMode mode = CompareMode.Flattened; //Mode of axis data comparison

Axis frontAxis = new Axis(frontAxisRecord, requiredProximity, mode);

When everything ready to create gesture model:

Axis frontAxis;
Axis sideAxis;
Axis vertAxis;
...
int id = 100; //Gesture identification number
long interval = 50 * 1000000; //Interval between measures in nanoseconds (50ms)
long cooldown = 1000 * 1000000; //Idleness interval after detection event in nanoseconds (1000ms)
long deviation = 200 * 1000000; //Possible deviation of total duration in nanoseconds (200ms)

DeviceGestureModel model = new DeviceGestureModel(id, frontAxis, sideAxis, vertAxis, interval, cooldown, deviation);

Then we may create detector object:

IGestureDetector detector = DeviceGestureLibrary.createGestureDetector(getApplicationContext());

For detection events we need listener:

class DetectListener implements IGestureDetectListener {

	public void onGestureDetected(int gestureID, long timestamp) {
	... //event processing
	}

}

Finally, activate gesture detection:

DeviceGestureModel model;
IGestureDetector detector;
DetectListener listener;
...
detector.registureGestureDetection(model, listener);

5. Example

---

Demo application on Google Play: https://play.google.com/store/apps/details?id=info.augury.gesturelibdemo

App usage video on Youtube: https://www.youtube.com/watch?v=8S83_gMkRnY

6. Hints

---

• Library can effectively detects very complex gestures, but only on precisely tuned models.
• If your gesture model do not use specific axis, it's highly recommended to set it null.
• If your gesture model have frequent direction shifts, ensure interval between measures is small enough.
• In rare cases, Straight compare mode is preferable rather than default Flattened mode.
• Complex gestures often require lower than average(0.7f) proximity threshold (at least on specific axis).
• Vice versa, common\short gestures require higher than average proximity threshold, to cut off false detection events.
• Often, there are sequence of detection events generated on single gesture - use model's cooldown parameter.