PCA: Invalid memory access when specifiying the dimensions

[palonso]

In PCA, featVector is defined as a matrix of size 1 x inputFeatureSize,

essentia/src/algorithms/highlevel/pca.cpp

Line 93 in 2de0185

Array2D<Real> featVector(1,bands, 0.0);

However, in this line featVector is implicitly resized to 1 x outputFeatureSize,

essentia/src/algorithms/highlevel/pca.cpp

Line 99 in 2de0185

featVector = matmult(featVector, reducedEig);

This causes an invalid memory access on the second iteration of the loop when inputFeatureSize != outputFeatureSize.

To reproduce the error set the parameter dimensions to anything smaller than the input feature size on any PCA routine.