search

twgo / siann1-hak8_boo5-hing5

聲學模型訓練
MIT License
1 stars 1 forks source link

8k , 8k a-law #41

Open sih4sing5hong5 opened 6 years ago

sih4sing5hong5 commented 6 years ago

tw01+tw02_8k

8k wav : sox -G /usr/local/gi2_liau7_khoo3/音檔/MH/MaternalHome-003.wav -b 16 -c 1 -r 16k -t wav - | avconv -i - -f wav -ar 8000 - |

tw01+tw02_8k a-law

| avconv -i - -f alaw -ar 8000 - | avconv -i - -f wav -ar 8000 -

仝語者_8k 仝語者_8k+a-alaw

tw01+tw02 tw01+tw02_8k tw01+tw02_8k_alaw

仝語者 仝語者_8k 仝語者_8k+a-alaw

leo424y commented 6 years ago

問到一個很酷的看不懂的模擬方式

先餵 LPF, cornet at 3.8kHz, attenuation > 60dB, downsample by 2. #聽不懂我就不解釋了

有時間會研究一下

leo424y commented 6 years ago

報告隊長，我加了 avconv 需要的套件 https://github.com/twgo/siann1-hak8_boo5-hing5/blob/5a50e1e6703a50eceb2625bed0ff99ba61633525/Dockerfile#L10

並依你的寫法在wav.scp每行尾巴多pipe執行avconv https://github.com/twgo/siann1-hak8_boo5-hing5/blob/5a50e1e6703a50eceb2625bed0ff99ba61633525/Dockerfile#L83

但似乎是無法轉的

我在想

我不應去改wav.scp

而是再它之前把wav壓過

讓wav.scp抓壓過的wav？

leo424y commented 6 years ago

試了

sox -G /usr/local/pian7sik4_gi2liau7/TW01/M0/TW01M0P0000/tbw1010.wav -b 16 -c 1 -r 16k -t wav - | avconv -i - -ar 8000 -

回應

Guessed Channel Layout for Input Stream #0.0 : mono Input #0, wav, from 'pipe:': Duration: N/A, bitrate: 256 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, 1 channels, s16, 256 kb/s [NULL @ 0xd09300] Unable to find a suitable output format for 'pipe:' pipe:: Invalid argument

看起來是參數要調整

sih4sing5hong5 commented 6 years ago

Unable to find a suitable output format for 'pipe:'

加 -f wav

Yang-Hsiang Chang notifications@github.com 於 2018年6月20日週三 上午10:21 寫道：

試了

sox -G /usr/local/pian7sik4_gi2liau7/TW01/M0/TW01M0P0000/tbw1010.wav -b 16 -c 1 -r 16k -t wav - | avconv -i - -ar 8000 -

回應

Guessed Channel Layout for Input Stream #0.0 : mono Input #0, wav, from 'pipe:': Duration: N/A, bitrate: 256 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, 1 channels, s16, 256 kb/s [NULL @ 0xd09300] Unable to find a suitable output format for 'pipe:' pipe:: Invalid argument

看起來是參數要調整

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/twgo/siann1-hak8_boo5-hing5/issues/41#issuecomment-398603253, or mute the thread https://github.com/notifications/unsubscribe-auth/AFuAC7sxNJN9q-yUptucn7YjnEnYd2ABks5t-bGcgaJpZM4Unrol .

leo424y commented 6 years ago

試了

/usr/local/kaldi/egs/taiwanese/s5c/data/train# sox -G /usr/local/pian7sik4_gi2liau7/TW01/M0/TW01M0P0000/tbw1010.wav -b 16 -c 1 -r 16k -t wav - | avconv -f wav -i - -ar 8000 -

回

Guessed Channel Layout for Input Stream #0.0 : mono Input #0, wav, from 'pipe:': Duration: N/A, bitrate: 256 kb/s Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 16000 Hz, 1 channels, s16, 256 kb/s [NULL @ 0x20d1320] Unable to find a suitable output format for 'pipe:' pipe:: Invalid argument

抱歉，我這樣直接改wav.scp對嗎？

RUN sed -i -z 's/\n/avconv -f wav -i - -ar 8000 - |\n/g' $KALDI_S5C/data/train/wav.scp

sih4sing5hong5 commented 6 years ago

你看一下 -f的參數，這是用佇output的，所以愛khng佇input後壁