Question

I’m trying to replicate the function seen here: http://en.wikipedia.org/wiki/Window_function#Blackman.E2.80.93Harris_window

But I simply can’t get any sense making values. This is my current code:

double blackman_harris(int n, int N){
double a0, a1, a2, a3, seg1, seg2, seg3, w_n;
a0 = 0.35875;
a1 = 0.48829;
a2 = 0.14128;
a3 = 0.01168;

seg1 = a1 * (double) cos((double)(2*M_PI*n)/(double) (N - 1));
seg2 = a2 * (double) cos((double)(4*M_PI*n)/(double) (N - 1));
seg3 = a3 * (double) cos((double)(6*M_PI*n)/(double) (N - 1));

w_n = a0 - seg1 + seg2 - seg3;

return w_n;
}

Thanks a ton for your help.

Answer 1

Define the window as a whole function.

bool VecBuildBlackmanHarrisWindow( float* pOut, unsigned int num )
{
    const float a0      = 0.35875f;
    const float a1      = 0.48829f;
    const float a2      = 0.14128f;
    const float a3      = 0.01168f;

    unsigned int idx    = 0;
    while( idx < num )
    {
        pOut[idx]   = a0 - (a1 * cosf( (2.0f * M_PI * idx) / (num - 1) )) + (a2 * cosf( (4.0f * M_PI * idx) / (num - 1) )) - (a3 * cosf( (6.0f * M_PI * idx) / (num - 1) ));
        idx++;
    }
    return true;
}

You can then define a window function as follows:

std::vector< float > window( 1024 );
VecBuildBlackmanHarrisWindow( &window.front(), window.size() );

Which means you can pre-calculate the window function in advance.

At this point I’m sorry I have led you wrongly. Sorry. I checked my code and you calculate the value by averaging all the window sample values together and then dividing by 2 (effectively add them all up and divide by N/2).

float fTotal    = 1.0f;
auto iter   = window.begin();
while( iter != window.end() )
{
    fTotal  += *iter;
    iter++;
}

fTotal  /= 1024.0f;
fTotal  /= 2.0f;

This gives me a value of 0.17969f

(If you use a larger window this value will change hence the value I gave you before being slightly lower).

Apologies for the confusion