自分のアプリを作ろう

0.690000	0.690000	S
8.360000	8.360000	S
12.410000	12.410000	S
14.040000	14.040000	S
48.930000	48.930000	S
139.470000	139.470000	S
148.060000	148.060000	S
197.890000	197.890000	S
235.860000	235.860000	S

import datetime
from pydub import AudioSegment

# 小分けファイルを作るための開始位置と終了位置のデータを返す
def SegBySilence():
    with open(AudacityLabel, "r") as f:
        Silence = f.readlines()
    i = 0
    segmentation = []
    for i in range(len(Silence)):
        try:
            j = i + 1
            StartSec = float(Silence[i].split("\t")[0])
            EndSec = float(Silence[j].split("\t")[0])

            # 長すぎ又は短すぎを防ぐ→Audacityに戻ってラベルの修正
            ClipLen = EndSec - StartSec
            td = datetime.timedelta(seconds=StartSec)
            if ClipLen > 10:      # 10秒を超える
                print("長すぎる！", td)
            if ClipLen < 1:
                print("短っ！", td)

            Duration = ("speech", StartSec, EndSec)
            segmentation.append(Duration)
        except IndexError:
            pass
    return segmentation

# セグメンテーションデータに基づき小分けファイルを作る
def Kowake(Segments):
    LoopCount = 0
    forJuliusSilence = AudioSegment.silent(duration=300)
    # 音声
    for Seg in Segments:
        # タプルの第1要素が区間のラベル
        SegLabel = Seg[0]

        if (SegLabel == 'speech'):  # 音声区間
            # 区間の開始時刻の単位を秒からミリ秒に変換
            StartSec  = Seg[1] * 1000
            EndSec = Seg[2] * 1000

            # 分割結果をwavに出力 Julius用
            NewAudio = AudioSegment.from_wav(AudioFile)
            NewAudio = forJuliusSilence + NewAudio[StartSec:EndSec] + forJuliusSilence

            OutFile = KowakeDir + "S" + str(LoopCount).zfill(4) + '.wav'
            OutVideo = KowakeDir + "V" + str(LoopCount).zfill(4) + '.mp4'

            NewAudio.export(OutFile, format="wav")

            # 分割結果をmp4に出力
            # ffmpeg -ss [開始地点(秒)] -i [入力する動画パス] -t [切り出す秒数] [出力する動画パス]
            Vstart = Seg[1]
            Vend = Seg[2]
            Sec = str(Vend - Vstart)
            StartTime = str(Vstart)
            cmd = "ffmpeg -ss " + StartTime + " -i " + VideoFile + " -t " + Sec + " " + OutVideo
            os.system(cmd)

            LoopCount += 1
            del NewAudio

# J か E でエンジンの種類を指定
Lang = "J"

def RunRecogEngine():
    os.chdir(KowakeDir)
    cmd = "ls -v S*.wav > wavlist.txt"      # ls -v natural sort of (version) numbers within text
    os.system(cmd)

    with open("wavlist.txt", "r") as f:
        Lines = f.readlines()

    if Lang == "J":
        cmd = "julius -C ../main.jconf -C ../am-dnn.jconf -quiet -dnnconf ../julius.dnnconf -input rawfile -filelist wavlist.txt -outfile -iwspword -rejectshort 300"
        os.system(cmd)
    elif Lang == "E":
        #cmd = "julius -C ../ENVR-v5.4.Dnn.Bin/julius.jconf -quiet -dnnconf ../ENVR-v5.4.Dnn.Bin/dnn.jconf -input rawfile -filelist wavlist.txt -outfile -iwspword -rejectshort 300"
        # Juliusに英語の音声モデルを指定してやってみたけどうまく動かなかった。
        for i in Lines:
            OutFile = i.split(".")[0] + ".out"
            cmd = "deepspeech --model ../DeepSpeech/deepspeech-0.7.0-models.pbmm --scorer ../DeepSpeech/deepspeech-0.7.0-models.scorer --audio " + i.strip() + " > " + OutFile
            os.system(cmd)
    else:
        exit()

    # .outファイルから必要な行だけ取ってEdit.Me（編集用全文ファイル)収録。
    EditMe = ""
    for l in Lines:
        JuliusOut = l.split(".")[0] + ".out"
        with open(JuliusOut, "r") as ff:
            Jout = ff.readlines()
        Jrecog = Jout[0]

        try:
            Jrecog = Jrecog.split("sentence1:  ")[1]
        except IndexError:
            pass

        # Juliusの出力には余計な文字が含まれる
        if Lang == "J":
            Jrecog = Jrecog.replace(" ", "")

        EditMe = EditMe + Jrecog

    if not os.path.exists(EditMeFile):
        with open(EditMeFile, "w") as fff:
            fff.write(EditMe)