【问题标题】:Sphinx 4 Failed to align audio to trancriptSphinx 4 无法将音频与脚本对齐
【发布时间】:2013-11-27 03:44:40
【问题描述】:

我正在使用 Sphinx 4 使用以下 wav files 跟踪声学模型适应。 这是我使用bw -hmmdir wsj -moddeffn wsj/mdef -ts2cbfn .cont. -feat 1s_c_d_dd -cmn current -agc none -dictfn vn.dic -ctlfn lisp.fileids -lsnfn lisp.transcription -accumdir . 时得到的结果

utt>     0                 lisp_0001   53INFO: cmn.c(175): CMN: 73.43  2.89 -0.3
4 -1.85 -0.98 -0.52  0.33  0.67 -0.77 -0.56  0.18 -0.50 -0.30
    0    28 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0001 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     1                 lisp_0002   41INFO: cmn.c(175): CMN: 74.39  2.48 -0.9
2 -2.09 -1.31 -0.52 -0.17  0.67  0.26 -0.62  0.34 -0.26 -0.04
    0    28 0 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0002 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     2                 lisp_0003   57INFO: cmn.c(175): CMN: 75.86  2.02 -0.5
3 -1.16 -0.79 -0.55 -0.77  0.92 -0.34 -0.82  0.63 -0.33 -0.60
    0    40 2 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0003 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     3                 lisp_0004   57INFO: cmn.c(175): CMN: 74.78  2.01 -0.3
6 -0.52 -1.04 -1.08 -0.08  0.88 -0.51 -0.65  0.56 -0.36 -0.54
    0    40 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0004 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     4                 lisp_0005   49INFO: cmn.c(175): CMN: 75.03  1.80 -1.7
6 -1.18 -1.56 -1.24  0.62  1.84 -0.58 -1.34  0.64 -0.26 -0.20
    0    28 2 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0005 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     5                 lisp_0006   41INFO: cmn.c(175): CMN: 76.75  0.51 -0.5
5 -0.89 -1.18 -1.16  0.64  1.67 -1.25 -1.30  0.57 -0.26 -0.54
    0    28 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0006 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     6                 lisp_0007   22INFO: cmn.c(175): CMN: 82.68 -5.14 -4.5
6 -1.20 -0.66 -0.34 -0.88 -0.05  1.29  1.60  0.97 -0.68 -1.65
    0    28 0 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0007 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     7                 lisp_0008   16INFO: cmn.c(175): CMN: 82.76 -6.31 -4.7
5 -1.98 -1.04 -1.06 -0.49  1.19  1.57  1.48  0.52 -1.17 -1.32
    0    28 0 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0008 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     8                 lisp_0009   47INFO: cmn.c(175): CMN: 78.49  1.93 -0.6
9 -0.95 -1.04 -0.06 -0.18  0.98 -0.98 -0.72  0.20  0.04 -0.54
    0    32 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0009 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>     9                 lisp_0010   47INFO: cmn.c(175): CMN: 77.21  1.23  0.1
8 -0.83 -0.89 -0.19 -0.39  0.80 -1.13 -0.86  0.38 -0.17 -0.47
    0    32 3 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0010 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    10                 lisp_0011   39INFO: cmn.c(175): CMN: 79.15  0.97  0.4
2 -0.53 -1.72 -1.64 -0.36  1.03  0.23 -0.49 -0.59 -0.21 -0.16
    0    32 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0011 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    11                 lisp_0012   41INFO: cmn.c(175): CMN: 77.22  1.29  0.4
5 -0.51 -2.12 -1.20 -0.52  1.09 -0.10 -0.56 -0.27 -0.60 -0.20
    0    36 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0012 ignored
 utt 0.038x 0.320e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    12                 lisp_0013   49INFO: cmn.c(175): CMN: 78.72  0.88 -0.8
3 -0.17 -0.09 -0.18 -1.40  0.71 -0.16 -1.00 -0.03  0.07 -0.35
    0    32 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0013 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    13                 lisp_0014   51INFO: cmn.c(175): CMN: 77.42  0.56 -0.8
8 -0.14 -0.05 -0.20 -1.31  0.90 -0.21 -1.39  0.07  0.01 -0.28
    0    32 3 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0014 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    14                 lisp_0015   57INFO: cmn.c(175): CMN: 74.21  1.50  0.0
8 -1.50 -1.63 -0.97  0.65  0.63 -0.30 -0.07 -0.25 -0.71 -0.21
    0    28 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0015 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    15                 lisp_0016   54INFO: cmn.c(175): CMN: 74.42  1.22  0.0
1 -1.77 -1.29 -1.20  0.30  0.83  0.39 -0.31 -0.32 -0.61 -0.11
    0    28 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0016 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    16                 lisp_0017   51INFO: cmn.c(175): CMN: 77.04  1.26 -1.0
4 -0.57 -0.62 -0.27 -0.04  0.25 -0.97 -0.66  0.42 -0.16 -0.32
    0    32 2 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0017 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    17                 lisp_0018   53INFO: cmn.c(175): CMN: 76.83  0.69 -1.3
5 -0.93 -0.46 -0.01 -0.53  0.61 -0.64 -1.03  0.85 -0.18 -0.15
    0    32 2 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0018 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    18                 lisp_0019   55INFO: cmn.c(175): CMN: 79.39  0.58 -0.5
1 -1.02 -1.71 -0.55  0.44  0.80  0.32 -0.67 -0.73 -0.09 -0.21
    0    36 2 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0019 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    19                 lisp_0020   53INFO: cmn.c(175): CMN: 77.16  1.12 -0.4
3 -1.27 -1.72 -1.32 -0.06  0.98  0.63 -0.42 -0.39 -0.03 -0.32
    0    32 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0020 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    20                 lisp_0023   43INFO: cmn.c(175): CMN: 78.04  1.22 -1.2
4 -1.15 -0.43 -0.20 -0.23  0.78 -0.33 -0.37  0.05 -0.60 -0.73
    0    24 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0023 ignored
 utt 0.036x 0.256e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    21                 lisp_0024   54INFO: cmn.c(175): CMN: 77.16  0.71 -1.2
7 -0.87 -0.57 -0.45  0.12  0.53  0.63 -0.43  0.26 -0.65 -0.38
    0    24 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0024 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    22                 lisp_0025   53INFO: cmn.c(175): CMN: 74.48  3.43  2.1
5 -0.18 -1.62 -0.61 -0.64 -0.19 -0.28  0.38  0.05 -0.40 -0.01
    0    32 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0025 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

utt>    23                 lisp_0026   33INFO: cmn.c(175): CMN: 69.14  5.30  2.1
9 -1.26 -2.55  0.65 -1.46  0.14 -0.22 -0.54  0.24 -0.34 -0.19
    0    32 1 ERROR: "backward.c", line 421: Failed to align audio to trancript:
 final state of the search is not reached
ERROR: "baum_welch.c", line 324: lisp_0026 ignored
 utt 0.000x 0.000e upd 0.000x 0.000e fwd 0.000x 0.000e bwd 0.000x 0.000e gau 0.0
00x 0.000e rsts 0.000x 0.000e rstf 0.000x 0.000e rstu 0.000x 0.000e

overall> stats 0 (-0) 0.000000e+000 0.000000e+000 0.000x 4.576e
WARNING: "accum.c", line 617: Over 500 senones never occur in the input data. Th
is is normal for context-dependent untied senone training or for adaptation, but
 could indicate a serious problem otherwise.
INFO: s3mixw_io.c(232): Wrote ./mixw_counts [4147x1x8 array]
INFO: s3tmat_io.c(174): Wrote ./tmat_counts [49x3x4 array]
INFO: s3gau_io.c(478): Wrote ./gauden_counts with means with vars [4147x1x8 vect
or arrays]
INFO: main.c(1014): Counts saved to .

我怀疑由于我的 WAV 文件没有任何静音垫,因此会发生上述错误。那是对的吗?如果不是,错误的原因是什么?

注意:我录制了一个包含所有单词的长音频文件,然后将它们剪切成单词。这是由于我的记录设备的保存速度慢。这会影响每个较小文件的质量吗?

提前致谢

【问题讨论】:

    标签: speech-recognition speech-to-text cmusphinx


    【解决方案1】:

    您的音频文件以 44.1khz 录制:

     file lisp_0009.wav
     lisp_0009.wav: RIFF (little-endian) data, WAVE audio, Microsoft PCM, 16 bit, mono 44100 Hz
    

    Sphinxtrain 需要 16khz 速率的音频,您可以使用 sox 重新采样音频:

     for f in *.wav; do 
         sox $f -r 16000 $f.new.wav; mv $f.new.wav $f; 
     done
    

    有关输入音频格式的更多信息,请参阅 CMUSphinx 适配教程

    http://cmusphinx.sourceforge.net/wiki/tutorialadapt

    我怀疑由于我的 WAV 文件没有任何静音垫,

    这也是一个问题。音频的边界必须有大约 0.25 秒的静默时间。

    然后将它们切成单词。

    你需要根据话语而不是单词来剪辑文件

    【讨论】:

      猜你喜欢
      • 1970-01-01
      • 1970-01-01
      • 2018-07-14
      • 1970-01-01
      • 2012-11-21
      • 2021-09-20
      • 1970-01-01
      • 1970-01-01
      • 2017-10-01
      相关资源
      最近更新 更多