An LDA/PLDA estimator using KALDI in python for speaker verification tasks
PLDA should be stable. So far only the targetdim function has some difficulties, but the general usage is possible