Pairwise distance between two matrices of 3 columns and 1911 rows.

Hello,

Hello.

I have two questions.

1)- I have two matrices of 3 columns and 1911 rows (matrices have similar dimensions).

In fact, each matrix consists of the data of a moving marker in 3D space. 1st column returns the amount of movement of the marker in the X vector (anteroposterior direction). 2nd column returns the amount of movement of the marker in the Y vector (vertical direction). 3rd column returns the amount of movement of the marker in the Z vector (mediolateral direction). The number of the rows is actually the number of frames that these 3 parameters have been recorded. I have to calculate the distance between each member of matrix A with the corresponding member of matrix B (e.g. 1st member of matrix A (which is the movement of the marker in the anteroposterior direction) with the 1st member of matrix B (which again is the movement of the marker in the anteroposterior direction). sqrt(sum((A - B) .^ 2)) just provides me with 7.6257 37.2261 5.2263 (for each column) while I have to calculate the distance of all 5733 (3*1911) memebrs of matrix A with corresponding member of matrix B.

2)- What I have to do if the number of rows of matrix A is not similar to the number of rows of matrix B?

I appreciate your kind attention to my request.

[madhan ravi]

Illustration with a shirt example matrix with expected result would help.

[Vahid Esmaeili]

I hope this illustration helps. As I mentioned, the first column consists of the data of position of the moving marker in anterioposterior direct, the 2nd column is its position in the vertical direction and the 3rd column is the position of the moving marker in the mediolateral direction. First marker is attached on the right ankle joint and the 2nd marker is attached on left ankle joint. I have to find the distance between each member of the first column (the anterioposterior position of marker of right ankle joint recorded in 1st frame) of matrix A (represented by the red circle) with the corresponding member of the first column (the anterioposterior position of marker of left ankle joint recorded in 1st frame) of matrix B., then the same distance between the 2nd memebers of 1st column of each matrix and I have to continue to the end. I have to do the same for the 2nd column (represented by the green circle) and 3rd column (represented by the pink circle) of matrix A with the corresponding member of 2nd and 3rd columns of matrix B.

[KSSV]

Read about pdist, pdist2.

[Vahid Esmaeili]

Thank you so much for your answer. It is not easy for me to understand and interpret the complicated output of 'pdist2'. I have to use the data in my thesis.

Answer 1

[Voss]

Sounds like you need to do the sum over the second dimension of (A-B).^2; that is, to sum the sqaured differences of positions in each of the X-, Y-, and Z-dimensions in space.

sqrt(sum((A - B) .^ 2,2))

See the documentation for sum for further clarification.

[Vahid Esmaeili]

Thank you so much for your reply.