ivector-plda-scoring-dense.cc File Reference

#include "base/kaldi-common.h"
#include "util/common-utils.h"
#include "util/stl-utils.h"
#include "ivector/plda.h"

Include dependency graph for ivector-plda-scoring-dense.cc:

Go to the source code of this file.

Namespaces
	kaldi
	This code computes Goodness of Pronunciation (GOP) and extracts phone-level pronunciation feature for mispronunciations detection tasks, the reference:

Functions
bool	EstPca (const Matrix< BaseFloat > &ivector_mat, BaseFloat target_energy, const std::string &reco, Matrix< BaseFloat > *mat)
void	TransformIvectors (const Matrix< BaseFloat > &ivectors_in, const PldaConfig &plda_config, const Plda &plda, Matrix< BaseFloat > *ivectors_out)
void	ApplyPca (const Matrix< BaseFloat > &ivectors_in, const Matrix< BaseFloat > &pca_mat, Matrix< BaseFloat > *ivectors_out)
int	main (int argc, char *argv[])

Function Documentation

main()

int main	(	int	argc,
		char *	argv[]
	)

Definition at line 114 of file ivector-plda-scoring-dense.cc.

References kaldi::ApplyPca(), Plda::ApplyTransform(), SequentialTableReader< Holder >::Done(), kaldi::EstPca(), ParseOptions::GetArg(), RandomAccessTableReader< Holder >::HasKey(), rnnlm::i, rnnlm::j, KALDI_ASSERT, KALDI_ERR, KALDI_LOG, KALDI_WARN, SequentialTableReader< Holder >::Key(), Plda::LogLikelihoodRatio(), SequentialTableReader< Holder >::Next(), ParseOptions::NumArgs(), ParseOptions::PrintUsage(), ParseOptions::Read(), kaldi::ReadKaldiObject(), PldaConfig::Register(), ParseOptions::Register(), MatrixBase< Real >::Row(), kaldi::TransformIvectors(), RandomAccessTableReader< Holder >::Value(), SequentialTableReader< Holder >::Value(), and TableWriter< Holder >::Write().

                                 {
  using namespace kaldi;
  typedef kaldi::int32 int32;
  try {
    const char *usage =
      "Perform PLDA scoring for speaker diarization.  The input reco2utt\n"
      "should be of the form <recording-id> <seg1> <seg2> ... <segN> and\n"
      "there should be one iVector for each segment.  PLDA scoring is\n"
      "performed between all pairs of iVectors in a recording and outputs\n"
      "an archive of score matrices, one for each recording-id.  The rows\n"
      "and columns of the the matrix correspond the sorted order of the\n"
      "segments.\n"
      "Usage: ivector-plda-scoring-dense [options] <plda> <reco2utt>"
      " <ivectors-rspecifier> <scores-wspecifier>\n"
      "e.g.: \n"
      "  ivector-plda-scoring-dense plda reco2utt scp:ivectors.scp"
      " ark:scores.ark ark,t:ivectors.1.ark\n";

    ParseOptions po(usage);
    BaseFloat target_energy = 0.5;
    PldaConfig plda_config;
    plda_config.Register(&po);

    po.Register("target-energy", &target_energy,
      "Reduce dimensionality of i-vectors using a recording-dependent"
      " PCA such that this fraction of the total energy remains.");
    KALDI_ASSERT(target_energy <= 1.0);

    po.Read(argc, argv);

    if (po.NumArgs() != 4) {
      po.PrintUsage();
      exit(1);
    }

    std::string plda_rxfilename = po.GetArg(1),
      reco2utt_rspecifier = po.GetArg(2),
      ivector_rspecifier = po.GetArg(3),
      scores_wspecifier = po.GetArg(4);

    Plda plda;
    ReadKaldiObject(plda_rxfilename, &plda);

    SequentialTokenVectorReader reco2utt_reader(reco2utt_rspecifier);
    RandomAccessBaseFloatVectorReader ivector_reader(ivector_rspecifier);
    BaseFloatMatrixWriter scores_writer(scores_wspecifier);
    int32 num_reco_err = 0,
          num_reco_done = 0;
    for (; !reco2utt_reader.Done(); reco2utt_reader.Next()) {
      Plda this_plda(plda);
      std::string reco = reco2utt_reader.Key();

      std::vector<std::string> uttlist = reco2utt_reader.Value();
      std::vector<Vector<BaseFloat> > ivectors;

      for (size_t i = 0; i < uttlist.size(); i++) {
        std::string utt = uttlist[i];

        if (!ivector_reader.HasKey(utt)) {
          KALDI_ERR << "No iVector present in input for utterance " << utt;
        }

        Vector<BaseFloat> ivector = ivector_reader.Value(utt);
        ivectors.push_back(ivector);
      }
      if (ivectors.size() == 0) {
        KALDI_WARN << "Not producing output for recording " << reco
                   << " since no segments had iVectors";
        num_reco_err++;
      } else {
        Matrix<BaseFloat> ivector_mat(ivectors.size(), ivectors[0].Dim()),
                          ivector_mat_pca,
                          ivector_mat_plda,
                          pca_transform,
                          scores(ivectors.size(), ivectors.size());

        for (size_t i = 0; i < ivectors.size(); i++) {
          ivector_mat.Row(i).CopyFromVec(ivectors[i]);
        }
        if (EstPca(ivector_mat, target_energy, reco, &pca_transform)) {
          // Apply the PCA transform to the raw i-vectors.
          ApplyPca(ivector_mat, pca_transform, &ivector_mat_pca);

          // Apply the PCA transform to the parameters of the PLDA model.
          this_plda.ApplyTransform(Matrix<double>(pca_transform));

          // Now transform the i-vectors using the reduced PLDA model.
          TransformIvectors(ivector_mat_pca, plda_config, this_plda,
            &ivector_mat_plda);
        } else {
          // If EstPca returns false, we won't apply any PCA.
          TransformIvectors(ivector_mat, plda_config, this_plda,
          &ivector_mat_plda);
        }
        for (int32 i = 0; i < ivector_mat_plda.NumRows(); i++) {
          for (int32 j = 0; j < ivector_mat_plda.NumRows(); j++) {
            scores(i, j) = this_plda.LogLikelihoodRatio(Vector<double>(
              ivector_mat_plda.Row(i)), 1.0,
              Vector<double>(ivector_mat_plda.Row(j)));
          }
        }
        scores_writer.Write(reco, scores);
        num_reco_done++;
      }
    }
    KALDI_LOG << "Processed " << num_reco_done << " recordings, "
              << num_reco_err << " had errors.";
    return (num_reco_done != 0 ? 0 : 1 );
  } catch(const std::exception &e) {
    std::cerr << e.what();
    return -1;
  }
}