Hi , I have a major question , it could be the last if it works ! 
Actually we have access to the pitch , but I also need the angle of attack, is it possible ?
You can calculate it roughly (EDIT for level flight) if you know
Airspeed V
The wing area A
mass of the aeroplane m
Gravity constant g
air density rho
where cl is the unknown lift coefficient
Lift = 1/2 * cl * rho * A * V^2
To sustain the aircraft in flight also
Lift = m * g
The angle of attack in degrees is roughly equal to cl/ 0.11 ref[1]
rearranging and substituting the formula (If my working is correct!)
where alpha_deg is the unknown angle of attack in degrees
alpha_deg = 2/0.11 * m * g / (rho * A * V^2)
For shallow angles involving pitch roll and yaw
you could probably start by working out the vertical projected wing area
ref[1]
www.srmuniv.ac.in/sites/default/files/downloads/class2-2012.pdf
Firstly thanks you a lot for helping me ! That’s a good idea, but there is still one problem. As I’m in climbing flight my equation becomes Lift=mgcos(slope) . And I have not the slope angle 
Or maybe you have used the approximation that mgcos(slope)=mg for small angle. (slope<<1) ?
And in an other hand, do you think there is a function to return directly alpha in Arduplane ?
The question could be, in arduplane, does slope and angle of attack exist ? Is there only the pitch parametered in the code ? because I 've found nothing about slope or angle of attack in the code ;/
Actually, skyscraper’s assumptions work regardless of climb angle. The only difference is that instead of “g” in the equations you use “a_z” (Z axis acceleration). I’ve used this assumption and validated it against an actual AoA probe (made by Aeroprobe) and the results agree to within a degree during flight.
You might need a different constant term, but the basic assumption (AoA proportional to some constant times “a_z” and “V^2”) holds during normal, non-stalled flight, in any attitude.
If you want slope, you need to differentiate position or combine velocity and climb rate. But slope isn’t a good way to calculate AoA compared to the method mentioned earlier.
I would have said that Ardupilot cant provide angle of attack since it doesn’t know the mass and wing area parameters ( also such things as aspect ratio or overall aerodynamic efficiency which has a big effect in practise which would give a more accurate result)
however there is currently a pull request to make an estimate of angle of attack available.
I think it may require an angle of attack sensor though
Okay thank you for your replies , I needed this angle to find the Vx valuers of my plane from the Aispeed. Finally I have decided to find it by deriving positions.