I have an app that reads the values returned by the magnetometer.
The way it does it at the moment is in the phone's coordinates system.
I want to be able to always read the magnetic field vector in a global coordinates system (x,y,z - east, north, sky)
The code I tried (inspired from this question) will result in a value of 0 for the x axis component, and variable y,z depending on the way I tilt the phone. As far I understood, the rotation matrix transform should make my coordinates system global, but it doesn't seem that way.
The expected behaviour would be to have the same value on the x,y,z components as long as I hold the phone in one place, no matter its tilt.
private final float alpha = (float) 0.8;
private float gravity[] = new float[3];
private float magnetic[] = new float[3];
public void onSensorChanged(SensorEvent event) {
Sensor sensor = event.sensor;
if (sensor.getType() == Sensor.TYPE_ACCELEROMETER) {
// Isolate the force of gravity with the low-pass filter.
gravity[0] = alpha * gravity[0] + (1 - alpha) * event.values[0];
gravity[1] = alpha * gravity[1] + (1 - alpha) * event.values[1];
gravity[2] = alpha * gravity[2] + (1 - alpha) * event.values[2];
} else if (sensor.getType() == Sensor.TYPE_MAGNETIC_FIELD) {
magnetic[0] = event.values[0];
magnetic[1] = event.values[1];
magnetic[2] = event.values[2];
float[] R = new float[9]; //rotation matrix
float[] I = new float[9]; //inclination
SensorManager.getRotationMatrix(R, I, gravity, magnetic);
float [] A_D = event.values.clone(); // device coordinates
float [] A_W = new float[3]; // global coodinates
A_W[0] = R[0] * A_D[0] + R[1] * A_D[1] + R[2] * A_D[2];
A_W[1] = R[3] * A_D[0] + R[4] * A_D[1] + R[5] * A_D[2];
A_W[2] = R[6] * A_D[0] + R[7] * A_D[1] + R[8] * A_D[2];
Log.d("Field","\nX :"+A_W[0]+"\nY :"+A_W[1]+"\nZ :"+A_W[2]);
}
The code is correct but there is always fluctuation as gravity is only an estimate. Actually, if the device is still, the accelerometer in the world coordinate should theoretically be the same independent of position since the only force acting on it is minus gravity and thus accelerometer should be gravity when the device is still. You can only expect that the East and North coordinates are very small when the device is still.
Related
How to get roll around Y axis?, I had achieved this when device is flat but not working when device hold on vertical, in front of face :( please help me with this.
float[] gravity = new float[3];
private float[] acceleration = new float[3];
//Above globel variables
if (event.sensor.getType() == Sensor.TYPE_ACCELEROMETER) {
lowPassFilter.filter(event.values);
System.arraycopy(acceleration, 0, this.acceleration, 0, acceleration.length);
gravity = event.values.clone();
double norm_Of_g = Math.sqrt(gravity[0] * gravity[0] + gravity[1] * gravity[1] + gravity[2] * gravity[2]);
// Normalize the accelerometer vector
gravity[0] = (float) (gravity[0] / norm_Of_g);
gravity[1] = (float) (gravity[1] / norm_Of_g);
gravity[2] = (float) (gravity[2] / norm_Of_g);
int rotation = (int) Math.round(Math.toDegrees(Math.atan2(gravity[0], gravity[1])));
tvXYZ.setText(String.format("X:%.2f, Y:%.2f, Z:%.2f\nArray Size: %d, Position in Array : %d",
gravity[0], gravity[1], gravity[2], list.size(), rotation));
imageView.setRotationY(rotation);
}
Have tried to adapt the android developer documentation for accelerator LPF but it does not seem to work with pressure
float pressure_value = 0.0f;
float height = 0.0f;
float height2 = (float) 964.98;
final float alpha = (float) 0.8;
if( Sensor.TYPE_PRESSURE == event.sensor.getType() ) {
pressure_value = event.values[0];
event.values[0] = alpha * event.values[0] + (1 - alpha) * event.values[0];
event.values[1] = alpha * event.values[1] + (1 - alpha) * event.values[1];
event.values[2] = alpha * event.values[2] + (1 - alpha) * event.values[2];
does anyone have some insight?
The pressure sensor only has one value i.e. event.values[0]. A simple low pass filter for that would look like:
pressure = alpha*event.values[0] + (1 - alpha)*pressure
Extremely oversimplified explanation:
For alpha=0.8, the 'new' pressure value is 80% of the actual current pressure supplied by the sensor + 20% of the value of the 'old' pressure. Increasing the alpha value will make it more responsive to pressure fluctuations, lower alpha values will make it less noisy (more filtered).
More explanatory code:
private float alpha = 0.8f;
private float filteredPressure = 0.0f;
#Override
public final void onSensorChanged(SensorEvent event) {
if (Sensor.TYPE_PRESSURE == event.sensor.getType()) {
float currentPressure = event.values[0];
filteredPressure = (alpha*currentPressure) + (1 - alpha)*filteredPressure;
}
}
I am trying to track the movement of the device only on the vertical direction, i.e. upward and downward movement. This should be irrespective of the orientation of the device. Things that i already know or have tried are these
Linear acceleration is given by sensor TYPE_LINEAR_ACCELERATION and the axes is the phone axes and hence tracking any particular axes does not make a difference.
I tried applying transpose or inverse of rotation vector( inverse or transpose for the rotation vector are same) and then tried tracking the z direction of the linear acceleration vector. Does not seem to help.
I am trying to do a dot product with gravity values (TYPE_GRAVITY) to get the direction of the acceleration but it seems to be error prone. Even when i move my device swiftly up, it says going down.
I will outline this method here
dotProduct = vectorA[0]*vectorB[0]+vectorA[1]*vectorB[1] + vectorA[2]*vectorB[2];
cosineVal = dotProduct/(|vectorA|*|vectorB|)
if(cosineVal > 0 ) down else Up.
What is the flaw with the method ? Please help, I have been stuck on this for some time now.
As I see it, in the 3rd method you trying to find the cos of angle between two vectors (gravity vector and acceleration vector). And the idea is if the angle is close to 180 degrees you have up movement, if angle is close to 0 degrees you have down movement. Cosine is function that has positive value when angle is from -90 to 90 degrees. So when your cosineVal value is positive it means phone is going down and even if cosineVal closer to 1 movement is straight down. So it is true vice versa. When cosine is negative ( from 90 degrees to 270) you have up movement.
Probably you can get vectors from Sensor.TYPE_ACCELEROMETER from https://developer.android.com/reference/android/hardware/SensorEvent.html#values there you have gravity vector and acceleration vector.
I made a code snippet below you can try.
public class MainActivity extends AppCompatActivity implements SensorEventListener {
private float[] gravity = new float[3];
private float[] linear_acceleration = new float[3];
#Override
protected void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
setContentView(R.layout.activity_main);
SensorManager mSensorManager = (SensorManager) getSystemService(SENSOR_SERVICE);
Sensor mAccelerometer = mSensorManager.getDefaultSensor(Sensor.TYPE_ACCELEROMETER);
mSensorManager.registerListener(this, mAccelerometer, SensorManager.SENSOR_DELAY_NORMAL);
}
#Override
public void onSensorChanged(SensorEvent event) {
// alpha is calculated as t / (t + dT)
// with t, the low-pass filter's time-constant
// and dT, the event delivery rate
final float alpha = 0.8f;
gravity[0] = alpha * gravity[0] + (1 - alpha) * event.values[0];
gravity[1] = alpha * gravity[1] + (1 - alpha) * event.values[1];
gravity[2] = alpha * gravity[2] + (1 - alpha) * event.values[2];
linear_acceleration[0] = event.values[0] - gravity[0];
linear_acceleration[1] = event.values[1] - gravity[1];
linear_acceleration[2] = event.values[2] - gravity[2];
float scalarProduct = gravity[0] * linear_acceleration[0] +
gravity[1] * linear_acceleration[1] +
gravity[2] * linear_acceleration[2];
float gravityVectorLength = (float) Math.sqrt(gravity[0] * gravity[0] +
gravity[1] * gravity[1] + gravity[2] * gravity[2]);
float lianearAccVectorLength = (float) Math.sqrt(linear_acceleration[0] * linear_acceleration[0] +
linear_acceleration[1] * linear_acceleration[1] + linear_acceleration[2] * linear_acceleration[2]);
float cosVectorAngle = scalarProduct / (gravityVectorLength * lianearAccVectorLength);
TextView tv = (TextView) findViewById(R.id.tv);
if (lianearAccVectorLength > 2) {//increase to detect only bigger accelerations, decrease to make detection more sensitive but noisy
if (cosVectorAngle > 0.5) {
tv.setText("Down");
} else if (cosVectorAngle < -0.5) {
tv.setText("Up");
}
}
}
#Override
public void onAccuracyChanged(Sensor sensor, int i) {
}
}
I came across this solution to remove gravity values from raw values of an acelerometer but I can't understand why alpha = 0.8 and not just 0.997.
public void onSensorChanged(SensorEvent event)
{
// alpha is calculated as t / (t + dT)
// with t, the low-pass filter's time-constant
// and dT, the event delivery rate
final float alpha = 0.8;
gravity[0] = alpha * gravity[0] + (1 - alpha) * event.values[0];
gravity[1] = alpha * gravity[1] + (1 - alpha) * event.values[1];
gravity[2] = alpha * gravity[2] + (1 - alpha) * event.values[2];
linear_acceleration[0] = event.values[0] - gravity[0];
linear_acceleration[1] = event.values[1] - gravity[1];
linear_acceleration[2] = event.values[2] - gravity[2];
}
As per the comment on your code: alpha = t / (t + dT), so alpha is just to be calculated according to your filter time lenght and the sampling rate of the accelerometer.
if your solution sets alpha = 0.8 it simply means that dT = 0,025 t or, on the other side, t = 40dt.
If you increase the sampling rate of the accelerometer, decreasing dT you'll have alpha pointing asyntotically to 1.
I tried the following way, but it doesn't work perfectly.. If device is kept in normal way, then also it fires shake event.
#Override
public void onSensorChanged(SensorEvent event) {
// TODO Auto-generated method stub
long curTime = System.currentTimeMillis();
// only allow one update every 200Ms.
if ((curTime - lastUpdate) > 200) {
lastUpdate = curTime;
x = event.values[SensorManager.DATA_X];
y = event.values[SensorManager.DATA_Y];
z = event.values[SensorManager.DATA_Z];
Vibrator vibrate = (Vibrator) getSystemService(Context.VIBRATOR_SERVICE);
if (Round(y, 4) > 22) {
vibrate.vibrate(200);
Log.d("sensor", "==== Up Detected===");
} else if (Round(y, 4) < -20) {
vibrate.vibrate(200);
Log.d("sensor", "==== Down Detected=== ");
}
mLastX = x;
mLastY = y;
mLastZ = z;
}
}
public static float Round(float Rval, int Rpl) {
float p = (float) Math.pow(10, Rpl);
Rval = Rval * p;
float tmp = Math.round(Rval);
return (float) tmp / p;
}
Please Help.
Any Help would be highly appreciated..
Thanks
Sensor values are very raw and noisy. You need to add a layer of digital signal processing on top of them to get good results. Just using the raw values will lead to a lot of jitter in your results. You're trying to do some basic ones with your time delay, but you need to do more filtering.
Also, you're vibrating in response to a shake. That vibrate will cause the accelerometer to see movement, leading to more false positives.
Apply lowpass filter on as your y-axis values.
The basic filter will be as mentioned in documentation.
public void onSensorChanged(SensorEvent event){
// In this example, alpha is calculated as t / (t + dT),
// where t is the low-pass filter's time-constant and
// dT is the event delivery rate.
final float alpha = 0.8;
// Isolate the force of gravity with the low-pass filter.
gravity[0] = alpha * gravity[0] + (1 - alpha) * event.values[0];
gravity[1] = alpha * gravity[1] + (1 - alpha) * event.values[1];
gravity[2] = alpha * gravity[2] + (1 - alpha) * event.values[2];
// Remove the gravity contribution with the high-pass filter.
linear_acceleration[0] = event.values[0] - gravity[0];
linear_acceleration[1] = event.values[1] - gravity[1];
linear_acceleration[2] = event.values[2] - gravity[2];
}