You could try to use a Least Squares Estimator (ls-estimator). It can fit a curve with unknown amplitude an phase to a signal. A least squares estimator always consists of an observation matrix. In this you can "design" your "cuttet sine wave" like this (if one period of your sine wave consists of 8 samples an your signal only contains one period):
x1 = [0 0.707 1 0.707 0 0 0 0]^T
To estimate the phase (and the true amplitude) you must fit a cosine wave, too:
x2 = [0 0 0 0 0 0.707 1 0.707]^T
So you observation matrix (to estimate amplitude and phase of one sinewave with know frequency) is:
X = [x1, x2]
The formular of the least square estimator is:
b = (X^T * X)^(-1) * X^Ty
with b containing the amplitdue of the sine and the cosine wave.
BTW ls-esitmator is quite powerfull an can be applied to many problems. It's worth to learn. Good look.
This paper paper shows the main ideas within the first two pages: