From the above information I'm assuming that data is a matrix of size 168x65. So data(:) returns a new array of size 10920x1 (all elements concatenated to a single column vector) and you are performing kmeans clustering on this new array and not on the actual data matrix.
Hence the kmeans is performed on the 10920 data points with each data point having dimension 1 forming 6 clusters and hence the cluster centers will be of dimension 1 as well.
If you are looking for your cluster center to be of dimension 65 then use the kmeans function as follows:
[idx c]=kmeans(data,6);
If this is not your issue, can you provide more information with relevant code and examples.
