Esempio n. 1
0
def lattice_1best(lattice,
                  *,
                  acoustic_scale=1.0,
                  lm_scale=1.0,
                  word_ins_penalty=0.0):
    args = kaldibin._magic_parse_args(lattice_1best, **locals())
    return kaldibin._context.run("src/latbin/lattice-1best",
                                 *args,
                                 wxtype="ark",
                                 wxfilename="-")
Esempio n. 2
0
def lattice_align_phones(model,
                         lattice,
                         *,
                         output_error_lats=True,
                         remove_epsilon=True,
                         reorder=True,
                         replace_output_symbols=False):
    args = kaldibin._magic_parse_args(lattice_align_phones, **locals())
    return kaldibin._context.run("src/latbin/lattice-align-phones",
                                 *args,
                                 wxtype="ark",
                                 wxfilename="-")
Esempio n. 3
0
def compute_cmvn_stats(feats,
                       *,
                       binary=True,
                       spk2utt='',
                       weights='',
                       wxtype='ark',
                       wxfilename='-'):
    args = kaldibin._magic_parse_args(compute_cmvn_stats, **locals())
    return kaldibin._context.run('src/featbin/compute-cmvn-stats',
                                 *args,
                                 wxtype=wxtype,
                                 wxfilename=wxfilename)
Esempio n. 4
0
def lattice_to_ctm_conf(lattice,
                        *,
                        acoustic_scale=1.0,
                        confidence_digits=2,
                        decode_mbr=True,
                        frame_shift=0.01,
                        inv_acoustic_scale=1.0,
                        lm_scale=1.0,
                        print_silence=False):
    args = kaldibin._magic_parse_args(lattice_to_ctm_conf, **locals())
    return kaldibin._context.run("src/latbin/lattice-to-ctm-conf",
                                 *args,
                                 wxtype=None,
                                 wxfilename="-")
Esempio n. 5
0
def lattice_determinize(lattice,
                        *,
                        acoustic_sale=1.0,
                        beam=10.0,
                        beam_ratio=0.9,
                        delta=0.000976562,
                        max_loop=500000,
                        max_mem=50000000,
                        minimize=False,
                        num_loops=20,
                        prune=False):
    args = kaldibin._magic_parse_args(lattice_determinize, **locals())
    return kaldibin._context.run("src/latbin/lattice-determinize",
                                 *args,
                                 wxtype="ark",
                                 wxfilename="-")
Esempio n. 6
0
def lattice_align_words_lexicon(lexicon,
                                model,
                                lattice,
                                *,
                                allow_duplicate_paths=False,
                                max_expand=-1,
                                output_error_lats=True,
                                output_if_empty=False,
                                partial_word_label=0,
                                reorder=True,
                                test=False):
    args = kaldibin._magic_parse_args(lattice_align_words_lexicon, **locals())
    return kaldibin._context.run("src/latbin/lattice-align-words-lexicon",
                                 *args,
                                 wxtype="ark",
                                 wxfilename="-")
Esempio n. 7
0
def compute_mfcc_feats(wavs,
                       *,
                       allow_downsample=False,
                       allow_upsample=False,
                       blackman_coeff=0.42,
                       cepstral_lifter=22,
                       channel=-1,
                       debug_mel=False,
                       dither=1.0,
                       energy_floor=0.0,
                       frame_length=25,
                       frame_shift=10,
                       high_freq=0,
                       htk_compat=False,
                       low_freq=20,
                       min_duration=0,
                       num_ceps=13,
                       num_mel_bins=23,
                       output_format='kaldi',
                       preemphasis_coefficient=0.97,
                       raw_energy=True,
                       remove_dc_offset=True,
                       round_to_power_of_two=True,
                       sample_frequency=16000,
                       snip_edges=True,
                       subtract_mean=False,
                       use_energy=True,
                       utt2spk='',
                       vtln_high=-500,
                       vtln_low=100,
                       vtln_map='',
                       vtln_warp=1,
                       window_type='povey',
                       verbose=0,
                       config='',
                       wxtype='ark',
                       wxfilename='-'):
    args = kaldibin._magic_parse_args(compute_mfcc_feats, **locals())
    return kaldibin._context.run('src/featbin/compute-mfcc-feats',
                                 *args,
                                 wxtype=wxtype,
                                 wxfilename=wxfilename)
Esempio n. 8
0
def nnet3_latgen_faster(nnet,
                        fst,
                        features,
                        *,
                        acoustic_scale=0.1,
                        allow_partial=False,
                        beam=16,
                        beam_delta=0.5,
                        computation_debug=False,
                        debug_computation=False,
                        delta=0.000976562,
                        determinize_lattice=True,
                        extra_left_context=0,
                        extra_left_context_initial=-1,
                        extra_right_context=0,
                        extra_right_context_final=-1,
                        frame_subsampling_factor=1,
                        frames_per_chunk=50,
                        hash_ratio=2.0,
                        ivectors='',
                        lattice_beam=10.0,
                        max_active=2147483647,
                        max_mem=50000000,
                        min_active=200,
                        minimize=False,
                        online_ivector_period=0,
                        online_ivectors='',
                        phone_determinize=True,
                        prune_interval=25,
                        utt2spk='',
                        word_determinize=True,
                        word_symbol_table='',
                        **args):
    '''
    WARNING: INCOMPLETE WRAPPER IMPLEMENTATION

    :param nnet:
    :param fst:
    :param features:
    :param _:
    :param acoustic_scale:
    :param allow_partial:
    :param beam:
    :param beam_delta:
    :param computation_debug:
    :param debug_computation:
    :param delta:
    :param determinize_lattice:
    :param extra_left_context:
    :param extra_left_context_initial:
    :param extra_right_context:
    :param extra_right_context_final:
    :param frame_subsampling_factor:
    :param frames_per_chunk:
    :param hash_ratio:
    :param ivectors:
    :param lattice_beam:
    :param max_active:
    :param max_mem:
    :param min_active:
    :param minimize:
    :param online_ivector_period:
    :param online_ivectors:
    :param phone_determinize:
    :param prune_interval:
    :param utt2spk:
    :param word_determinize:
    :param word_symbol_table:
    :param args:
    :return:
    '''

    args = kaldibin._magic_parse_args(nnet3_latgen_faster, **locals())
    return kaldibin._context.run('src/nnet3bin/nnet3-latgen-faster',
                                 *args,
                                 wxtype="ark",
                                 wxfilename="-")
Esempio n. 9
0
def nbest_to_ctm(nbest, *, frame_shift=0.01, precision=2, print_silence=False):
    args = kaldibin._magic_parse_args(nbest_to_ctm, **locals())
    return kaldibin._context.run("src/latbin/nbest-to-ctm",
                                 *args,
                                 wxtype=None,
                                 wxfilename="-")
Esempio n. 10
0
def wav_copy(wav, *, wxtype='ark', wxfilename='-'):
    args = kaldibin._magic_parse_args(compute_mfcc_feats, **locals())
    return kaldibin._context.run('src/featbin/wav-copy',
                                 *args,
                                 wxtype=wxtype,
                                 wxfilename=wxfilename)