Average or mean value of fixed-point array
c = mean(a)
c = mean(a,dim)
c = mean(a)a along
its first nonsingleton dimension.
c = mean(a,dim)a along
dimension dim. dim must
be a positive, real-valued integer with a power-of-two slope and a
bias of 0.
The input to the mean function must be a
real-valued fixed-point array.
The fixed-point output array c has
the same numerictype properties as the fixed-point
input array a. If the input, a,
has a local fimath, then it is used for intermediate
calculations. The output, c, is always
associated with the default fimath.
When a is an empty fixed-point array
(value = []), the value of the output array is
zero.
Compute the mean value along the first dimension (rows) of a fixed-point array.
x = fi([0 1 2; 3 4 5], 1, 32); % x is a signed FI object with a 32-bit word length % and a best-precision fraction length of 28-bits mx1 = mean(x,1)
Compute the mean value along the second dimension (columns) of a fixed-point array.
x = fi([0 1 2; 3 4 5], 1, 32); % x is a signed FI object with a 32-bit word length % and a best-precision fraction length of 28 bits mx2 = mean(x,2)
The general equation for computing the mean of
an array a, across dimension dim is:
sum(a,dim)/size(a,dim)
Because size(a,dim) is always a positive
integer, the algorithm casts size(a,dim) to an
unsigned 32-bit fi object with a fraction length
of zero (SizeA). The algorithm then computes the
mean of a according to the following equation,
where Tx represents the numerictype properties
of the fixed-point input array a:
c = Tx.divide(sum(a,dim), SizeA)