次へ: Noisy Speech Recogniton Using
上へ: Complex Spectrum Circle Centroid
戻る: Theoretical Properties of CSCC
It is obvious that the target signal spectrum is restored by
finding the centroid of the circle on which three or more microphone
inputs
lie. In the case of , the circle centroid is
uniquely determined from three distinct points on the circle. In the
case of microphone inputs, the circle centroid can be determined
as a point of nearly equal distance from observed microphone inputs. We
estimate the centroid as a point
by minimizing
the variance of squared distances from
, i.e.,
|
|
|
(3) |
where is a point on the complex spectrum
plane for arbitrary. We can include cases of or where the
minimum variance is 0.
To solve this equation, let and be real and imaginary parts of
, i.e., , and let and be those of
. Then, we have
from which its partial differentials in respect to and is
derived as:
denoting the covariance of and by
.
Letting the lefthand side of the above equation be 0, we obtain a linear
equation to obtain the centroid that minimizes the variance of squared
distances in Eq. (3):
|
(6) |
whose solutions and give the estimated complex spectrum centroid
for each frequency as
.
Eq. (6) has a solution if the covariance matrix:
|
(7) |
between and is regular.
Since its determinant is given by
|
(8) |
where is the correlation coefficient between and ,
the solution of Eq. (6) is guaranteed to exist unless
, i.e., all spectrum points
lie on a line in the complex plane.
Even though is always guaranteed to be no greater than 1, in
numerically bad conditions such as
, we use the center of
gravity of points
, i.e., the delay-and-sum solution,
instead of the circle centroid here.
次へ: Noisy Speech Recogniton Using
上へ: Complex Spectrum Circle Centroid
戻る: Theoretical Properties of CSCC
平成16年9月23日