Question

I’ve been given a MATLAB function that takes in a 1D vector and two sizes, the function then splits the data into blocks and finally stores them inside a 2D vector. I have been writing a C++ version of this function, but the (incorrect) results of my C++ function do not match the (correct) results of the MATLAB function.

Matlab function:

function f = block(v, N, M)

% This function separates the vector
% into blocks.  Each block has size N.
% and consecutive blocks differ in
% their starting positions by M
%
% Typically
%   N = 30 msec (600 samples)
%   M = 10 msec (200 samples)

  n = length(v);
  maxblockstart = n - N + 1;
  lastblockstart = maxblockstart - mod(maxblockstart-1 , M);

  % Remove the semicolon to see the number of blocks
  % numblocks = (lastblockstart-1)/M + 1
  numblocks = (lastblockstart-1)/M + 1;

  %f = zeros(numblocks,N);

  for i = 1:numblocks
  for j = 1:N
   f(i,j) = ((i-1)*M+j);
  end 
 end

For the purpose of this example, I’m just outputting the results of ((i-1)*M+j) and in MatLab I get these results (example):

1 201 401 601 .. 1001 1201 .. 1401 .. 1601 .. 1801

And here is my C++ function:

vector<iniMatrix> Audio::subBlocks(vector<float>& theData, int N, int M)
{
    // This method splits the vector into blocks
    // Each block has size N.
    // and consecutive blocks differ 
    
    int n = theData.size();
    int maxblockstart = n - N+1;
    int lastblockstart = maxblockstart - mod(maxblockstart-1, M);
    
    int numblocks = (lastblockstart-1)/M + 1;
    
    vector<float> subBlock;
    vector<iniMatrix> block;
    
    for(int i=1; (i < numblocks); i++)
    {
        for(int j=1; (j < N); j++)
        {
            cout << ((i-1)*M+j);
        }
        
    }

    return block;
}

The result I get from this:

1 2 3 4 .. 7 8 9 .. 13 14 15 etc..

P.S.

iniMatrix is just a typdef for the a vector of floats..

Another note, the variables:

n

maxblockstart

lastblockstart

numblocks

All have the same value in the Matlab program and the C++ so I think it’s something to do with the for loops..

Anyone have any suggestions?

Answer 1

Let me see if I’ve understood your desired algorithm correctly.
Say
n = 10
N = 3
M = 2

This should yield
0, 2, 4, 6 right? (since 8 won’t fit because 8+M is outside the range 0 <= x < n).

So, maxblockstart should be 7 = 10 – 3, that is n – N

lastblockstart should be 6 = 7 – 7 % 2, that is maxblockstart – maxblockstart % M

and numblocks should be 4 = 6/2 + 1, that is lastblockstart/M + 1

Modifying your code as follows, seems to yield the right results (only worked this out on paper, haven’t tried compiling or executing…..):

vector<iniMatrix> Audio::subBlocks(vector<float>& theData, int N, int M)
{
  int n = theData.size();
  int maxblockstart = n - N;
  int lastblockstart = maxblockstart - (maxblockstart % M);

  int numblocks = (lastblockstart)/M + 1;

  vector<iniMatrix> block;    

  for(int i=0; (i < numblocks); i++)
  {
     vector<float> subBlock;
      for(int j=0; (j < N); j++)
      {
          subBlock.push_back(theData[i*M+j]); //cout << (i*M+j);
      }  
      block.push_back(subBlock);
  }

  return block;
}

Give it a try…

Note that comparing results against MATLAB can be confusing, since C++ indicing is zero-based. Thus try the following;

1) Change the line cout << (i*M+j); to cout << theData[i*M+j];

2) Try the following test:

vector<float> test;
for(int i=0; i<=10000; i++)
  test.push_back(i);

Audio::subBlocks(test, 1023, 200);