speech_recognition实现录音ffmpeg实现音频文件转换，并用百度语音的sdk实现语音识别

版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/dss_dssssd/article/details/83540264

项目说明：

在windows平台下，使用speech_recognition记录音频，并转换为16k的wav， 之后利用ffmpeg将wav转化为pcm文件，上传到百度语音端，返回语音信息，并利用pyttsx3添加了简单的交互功能。

需求模块：

speech_recognition， pyttsx3， pyaudio， wave， aip， ffmpeg

模块安装：

• speech_recognition: https://pypi.org/project/SpeechRecognition/
• pyttsx3: https://blog.csdn.net/dss_dssssd/article/details/82693742
• pyaudio: https://pypi.org/project/PyAudio/
• aip:https://ai.baidu.com/docs#/ASR-Online-Python-SDK/top
• ffmpeg (Windows下） 注意是系统的环境变量，不是个人的path
  https://blog.csdn.net/zhuiqiuk/article/details/72834385

代码如下

import speech_recognition as sr
import pyttsx3
import pyaudio
import wave
from aip import AipSpeech
import os

# 读取wav文件并播放
def read_wav():
    CHUNK = 1024
# 测试语音
    wf = wave.open('./2.wav', 'rb')
        # read data
    data = wf.readframes(CHUNK)
    p = pyaudio.PyAudio()
    FORMAT = p.get_format_from_width(wf.getsampwidth())
    CHANNELS = wf.getnchannels()
    RATE = wf.getframerate()

    print('FORMAT: {} \nCHANNELS: {} \nRATE: {}'.format(FORMAT, CHANNELS, RATE))
    stream = p.open(format=FORMAT,

                    channels=CHANNELS,
                    rate=RATE,
                    frames_per_buffer=CHUNK,
                    output=True)
    # play stream (3)
    while len(data) > 0:
        stream.write(data)
        data = wf.readframes(CHUNK)



def wav_to_pcm(wav_file):
    # 假设 wav_file = "音频文件.wav"
    # wav_file.split(".") 得到["音频文件","wav"] 拿出第一个结果"音频文件"  与 ".pcm" 拼接 等到结果 "音频文件.pcm"
    pcm_file = "%s.pcm" %(wav_file.split(".")[0])
    # 就是此前我们在cmd窗口中输入命令,这里面就是在让Python帮我们在cmd中执行命令
    os.system("ffmpeg -y  -i %s  -acodec pcm_s16le -f s16le -ac 1 -ar 16000 %s"%(wav_file,pcm_file))

    return pcm_file

def get_file_content(filePath):
    with open(filePath, 'rb') as fp:
        return fp.read()


""" 你的 APPID AK SK """
# 需要根据自己申请的填写
# APP_ID = '你的 App ID'
# API_KEY = '你的 Api Key'
# SECRET_KEY = '你的 Secret Key'
# 这是测试id，key
APP_ID = '14545668'
API_KEY = 'BLG4GIxozxXia9U8KKtLBl2j'
SECRET_KEY = 'z0ITqlx8OXiveTePBvD7jkSCdGKthZAy'

def speech_interaction():
    # 初始化pyttsx3 engine
    engine = pyttsx3.init()

    # obtain audio from the microphone
    # 从麦克风记录数据
    r = sr.Recognizer()
    with sr.Microphone() as source:
        # print("Say something!")
        engine.say("门外有客人来访，需要开门吗, 请一秒后回答？")
        engine.runAndWait()
        r.adjust_for_ambient_noise(source)
        audio = r.listen(source)

    engine.say("录音结束, 识别中")
    engine.runAndWait()
    # 将数据保存到wav文件中
    with open("2.wav", "wb") as f: 
        f.write(audio.get_wav_data(convert_rate=16000))
     
     # 将记录的语音播放出来
     read_wav()      
    # 创建百度语音识别客户端
    client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)

    # 转成pcm格式
    pcmFile = wav_to_pcm("./2.wav")
    result = client.asr(get_file_content(pcmFile), 'pcm', 16000, {
        'dev_pid': 1537,
    })
    print(result)
    # print(result['err_msg'], result['result'][0])

    # 上传到百度云识别
    try:
        success = True if result['err_msg'] == 'success.' else False
        print(success)
        if success:
            text = result['result'][0]
            if "不" in text :
                engine.say("好的，那请您自己去开门")
                engine.runAndWait()
            elif "开" in text or '好' in text:
                engine.say("请您稍等，我去帮您开门，")
                engine.runAndWait()
            else:
                engine.say("语音识别错误")
                engine.runAndWait()
            # engine.say(text)
            # engine.runAndWait()
    except Exception as e:
        engine.say("抱歉， 识别错误")
        engine.runAndWait()

# 执行代码
speech_interaction()

注意：

pyttsx3的pyttsx3.engine()初始化不能放在线程中进行，会错。

说明：

• 如果返回timeout错误，在网络畅通的情况下，建议换一个id和key试一下。

项目放在github上了:
https://github.com/MengRe/speech_commmunication/tree/master