# 引入Speech SDK
from aip import AipSpeech
import subprocess
import datetime
import sys
import os
import time
from pydub import AudioSegment
import math
 
# 定義常量
#APP_ID = '你的 App ID'
APP_ID = '937****'
#API_KEY = '你的 API Key'
API_KEY = 'mOV9QaabNnkur0Aba15T****'
#SECRET_KEY = '你的 Secret Key'
SECRET_KEY = '097111374ad26d4ba00937c5e332****'
# 初始化AipSpeech對(duì)象
aipSpeech = AipSpeech(APP_ID, API_KEY, SECRET_KEY)
 
# 文件處理
def get_wave_filename(fileFullName):
 # MP3文件轉(zhuǎn)換成wav文件
 # 判斷文件后綴，是mp3的，直接處理為16k采樣率的wav文件；
 # 是wav的，判斷文件的采樣率，不是8k或者16k的，直接處理為16k的采樣率的wav文件
 # 其他情況，就直接返回AudioSegment直接處理
 fileSufix = fileFullName[fileFullName.rfind('.')+1:]
 print(fileSufix)
 filePath = fileFullName[:fileFullName.find(os.sep)+1]
 print(filePath)
 if fileSufix.lower() == "mp3":
 wavFile = "wav_%s.wav" %datetime.datetime.now().strftime('%Y%m%d%H%M%S')
 wavFile = filePath + wavFile
 cmdLine = "ffmpeg -i \"%s\" -ar 16000 " %fileFullName
 cmdLine = cmdLine + "\"%s\"" %wavFile
 print(cmdLine)
 ret = subprocess.run(cmdLine)
 print("ret code:%i" %ret.returncode)
 return wavFile
 #if ret.returncode == 1:
 # return wavFile
 #else:
 # return None
 else:
 return fileFullName
 
 
#文件分片
try:
 script, fileFullName = sys.argv
except:
 print("參數(shù) 文件名 未指定!")
 exit()
 
if not os.path.isfile(fileFullName):
 print("參數(shù) %s 不是一個(gè)文件名" %fileFullName)
 exit()
 
if not os.path.exists(fileFullName):
 print("參數(shù) %s 指定的文件不存在" %fileFullName)
 exit()
 
filePath = fileFullName[:fileFullName.find(os.sep)+1]
# 文件處理為Wav，采樣率16k的文件，返回文件名
wavFile = get_wave_filename(fileFullName)
print(wavFile)
record = AudioSegment.from_wav(wavFile)
if wavFile != fileFullName:
 time.sleep(1)
 os.remove(wavFile)
 
recLen = record.duration_seconds
interval = 120 * 1000
maxLoop = math.ceil(recLen*1000/float(interval))
for n in range(0,math.ceil(recLen*1000/float(interval))):
 recSeg = record[n * interval : (n + 1)*interval]
 #print("Segment:%i,startat:%i,length:%i" %n,n*interval/1000,recSeg.duration_seconds)
 print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + " >> Segment:" + str(n) +"/" + str(maxLoop))
 segFile = filePath + "seg%s.wav" %("0"*7 + str(n))[-6:]
 # 把分段的語(yǔ)音信息保存為臨時(shí)文件
 file_handle = recSeg.export(segFile,format="wav",codec = "libvorbis")
 file_handle.close()
 # 讀取分段的臨時(shí)文件為字節(jié)
 file_handle = open(segFile, 'rb')
 file_content = file_handle.read()
 file_handle.close()
 # 刪除臨時(shí)文件
 os.remove(segFile)
 # 用百度API處理該語(yǔ)音
 result=aipSpeech.asr(file_content, 'pcm', 16000, {'lan': 'zh'})
 if result['err_no'] == 0:
 print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + " >> " + result['result'][0])
 else:
 print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S') + " >> " + "err_no:" + str(result['err_no']))