Several of the multivariate operations require that the covariance matrix be inverted when calculating the output. If a covariance matrix cannot be inverted, it is considered singular. A matrix cannot be inverted if its determinant equals 0. That is, matrix A is invertible if there exists a matrix B such that AB = I = BA.
Check the values of your input raster bands to see if some of the bands are identical. Also check whether one or more bands contain all zeroes. If so, the specified classes may have some abnormalities. You should revisit your classification.