I am on a project to track the human body motion in to a 3d design and animation software. I found that acccelerometers are used for these purposes. According to my project what I will need is the path coordinates on which the human hand or other part moves. This can be done if I get the speed and integrate them i will find the path. But accelerometers by name seems that they gives the acceleration of the object they are attached, that is an object at rest will output zero and at shake time output some value and at motion time then output zero. So my question is this that does accelerometers outputs the accceleration of object or the force applied on them at a direction? If it is the force applied on a direction then it can work for me.
Accelerometers measure acceleration. Integrate twice for position with respect to time. Doing that for each axis provides a 3D position.
In theory. For the real world it's not even close to that simple. Doing a simple double integration will be reliable for less than a few seconds. If you want to find out a better way, then you'll need to ask a different question. Though you'll likely find it answered on this site already.
Typically accelerometers are not used for this because it's so unreliable. What people usually use are cameras and motion capture.
I am currently working with ADXL335 accelerometer. Basically, it is an electronic accelerometer working on x,y,z axes. It's output is a voltage signal. On it's datasheet, it is written that this ADXL335 accelerometer has a constant voltage of approx. 1.5 Volts, while it is stationary. So, this model can give you signal value even though it is not in motion. And this claim is tested and proved by me, I get a constant voltage value of 1.5 volts while it is stationary. While it is accelerated (I measured the vibration of a synchronous motor), it changes it's output voltage signal value by approximately 350 millivolts, so you can sense the acceleration. Maybe you can work with this, or a similar type of accelerometer.
First let me explain how an accelerometer measures what it does, as it seems that it is maybe not quite clear.
An accelerometer measures acceleration, hence the name. Most accelerometers have three sensors in total aligned to each of the axes x, y and z.
The actual data that comes from the sensor could be an analog voltage signal. The voltage will correspond to a normalized output and will depend on the hardware you use. It might for example output 1.5v when subjected to an acceleration of 10m/s² (~1G). Or it might include A2D output so that you basically get a digital number (say 1024) for that 1.5V signal. You would have to consult the docs for your particular hardware to see how the data scales to actual units, but there should be a nice linear relationship.
You can alternatively think of an accelerometer as a G force sensor.
If you leave the accelerometer undisturbed on a level table, you should get a normal reading on the z-axis only. Lets say that a normal reading is 1.5V or 1024 on a digital accelerometer. When you shove it in another direction, it will momentarily register that acceleration on the axis (or axes) that you shoved it in. It will only register zero on all axes when it is in free fall.
As an example, if you have an android phone download the "Accelerometer Monitor" application, there is likely a similar app for apple devices. It will show a graph of the sensor in each axis (x,y,z) as well a a total acceleration (|V|). Each axis will be a negative or positive number because the acceleration could be to the right or to the left (for the x axis), or forwards or backwards (for the y axis). The z axis will read a value at rest because of gravity which is producing a constant downward accelerating force. The |V| value is simply the square root of the squares of the x,y,z axes and |V| is the total G-force. It is always positive and represents the total acceleration of the sensor. The actual direction can be determined by analyzing the axes individually.
|V| = √(x²+y²+z²)
If you hold the sensor at an angle you will get a different reading on each of the three axes but |V| will still be the normal value (1.5v or 1024) if it is held steadily. Any motion of the accelerometer will lead to instantaneous readings on each axis that are proportional to the amount of extra force applied. Remember that once a body is in motion it will continue moving without a force due to inertia. So the accelerometer could be in your car going 100km/h and still only read 1G downwards.
Because it is only good for instantaneous readings, it is difficult to get accurate data that can be integrated over time, because its difficult to know what accelerations it was subject to between readings. You can guess and interpolate but the accumulation of error can compound quite quickly, which is why Samuel is suggesting that results can be unreliable.
To answer the question then, "what is the unit of measure?"
It will be some arbitrary unit depending on the sensor, but it can be scaled in a simple linear manner to either m/s² or to Gs (or to any other distance-unit per time-unit squared). 1G on Earth, at sea-level at the equator on a 20°C day, will be 9.80665 m/s².