轉自：https://blog.csdn.net/wolfblood_zzx/article/details/46418635

（百度的rest介面的部分網址發生了一定的變化，相關程式碼已更新） 

1、語音識別與語音合成的呼叫

註冊開發者帳號和建立應用的過程就不再贅述，百度的REST API在呼叫過程基本分為三步：

1. 獲取token
2. 向Rest介面提交資料
3. 處理返回資料

具體程式碼如下所示：

#!/usr/bin/python3

import urllib.request
import urllib
import json
import base64
class BaiduRest:
    def __init__(self, cu_id, api_key, api_secert):
        # token認證的url
        self.token_url = "https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id=%s&client_secret=%s"
 

        # 語音合成的resturl
        self.getvoice_url = "http://tsn.baidu.com/text2audio?tex=%s&lan=zh&cuid=%s&ctp=1&tok=%s"
        # 語音識別的resturl
        self.upvoice_url = 'http://vop.baidu.com/server_api'

        self.cu_id = cu_id
        self.getToken(api_key, api_secert)
        return
 


    def getToken(self, api_key, api_secert):
        # 1.獲取token
        token_url = self.token_url % (api_key,api_secert)

        r_str = urllib.request.urlopen(token_url).read()
        token_data = json.loads(r_str)
        self.token_str = token_data['access_token']
        pass

    def getVoice(self, text, filename):
        # 2. 向Rest介面提交資料
        get_url = self.getvoice_url % (urllib.parse.quote(text), self.cu_id, self.token_str)

        voice_data = urllib.request.urlopen(get_url).read()
        # 3.處理返回資料
        voice_fp = open(filename,'wb+')
        voice_fp.write(voice_data)
        voice_fp.close()
        pass

    def getText(self, filename):
        # 2. 向Rest介面提交資料
        data = {}
        # 語音的一些引數
        data['format'] = 'wav'
        data['rate'] = 8000
        data['channel'] = 1
        data['cuid'] = self.cu_id
        data['token'] = self.token_str
        wav_fp = open(filename,'rb')
        voice_data = wav_fp.read()
        data['len'] = len(voice_data)
        data['speech'] = base64.b64encode(voice_data).decode('utf-8')
        post_data = json.dumps(data)
        r_data = urllib.request.urlopen(self.upvoice_url,data=bytes(post_data,encoding="utf-8")).read()
        # 3.處理返回資料
        return json.loads(r_data)['result']

if __name__ == "__main__":
    # 我的api_key,供大家測試用，在實際工程中請換成自己申請的應用的key和secert
    api_key = "SrhYKqzl3SE1URnAEuZ0FKdT" 
    api_secert = "hGqeCkaMPb0ELMqtRGc2VjWdmjo7T89d"
    # 初始化
    bdr = BaiduRest("test_python", api_key, api_secert)
    # 將字串語音合成並儲存為out.mp3
    bdr.getVoice("你好！恭喜你的比原鏈又漲了!", "out.mp3")
    # 識別test.wav語音內容並顯示
    print(bdr.getText("out.wav"))

2、呼叫pyaudio使用麥克風錄製聲音

python中的pyaudio庫可以直接通過麥克風錄製聲音，可使用pip進行安裝。我們可以通過呼叫該庫，獲取到wav測試語音。 
具體程式碼如下所示：

#!/usr/bin/python3
# -*- coding: utf-8 -*-

from pyaudio import PyAudio, paInt16 
import numpy as np 
from datetime import datetime 
import wave

class recoder:
    NUM_SAMPLES = 2000      #pyaudio內建緩衝大小
    SAMPLING_RATE = 8000    #取樣頻率
    LEVEL = 500         #聲音儲存的閾值
    COUNT_NUM = 20      #NUM_SAMPLES個取樣之內出現COUNT_NUM個大於LEVEL的取樣則記錄聲音
    SAVE_LENGTH = 8         #聲音記錄的最小長度：SAVE_LENGTH * NUM_SAMPLES 個取樣
    TIME_COUNT = 60     #錄音時間，單位s

    Voice_String = []

    def savewav(self,filename):
        wf = wave.open(filename, 'wb') 
        wf.setnchannels(1) 
        wf.setsampwidth(2) 
        wf.setframerate(self.SAMPLING_RATE) 
        wf.writeframes(np.array(self.Voice_String).tostring()) 
        # wf.writeframes(self.Voice_String.decode())
        wf.close() 

    def recoder(self):
        pa = PyAudio() 
        stream = pa.open(format=paInt16, channels=1, rate=self.SAMPLING_RATE, input=True, 
            frames_per_buffer=self.NUM_SAMPLES) 
        save_count = 0 
        save_buffer = [] 
        time_count = self.TIME_COUNT

        while True:
            time_count -= 1
            # print time_count
            # 讀入NUM_SAMPLES個取樣
            string_audio_data = stream.read(self.NUM_SAMPLES) 
            # 將讀入的資料轉換為陣列
            audio_data = np.fromstring(string_audio_data, dtype=np.short)
            # 計算大於LEVEL的取樣的個數
            large_sample_count = np.sum( audio_data > self.LEVEL )
            print(np.max(audio_data))
            # 如果個數大於COUNT_NUM，則至少儲存SAVE_LENGTH個塊
            if large_sample_count > self.COUNT_NUM:
                save_count = self.SAVE_LENGTH 
            else: 
                save_count -= 1

            if save_count < 0:
                save_count = 0 

            if save_count > 0 : 
            # 將要儲存的資料存放到save_buffer中
                #print  save_count > 0 and time_count >0
                save_buffer.append( string_audio_data ) 
            else: 
            #print save_buffer
            # 將save_buffer中的資料寫入WAV檔案，WAV檔案的檔名是儲存的時刻
                #print "debug"
                if len(save_buffer) > 0 : 
                    self.Voice_String = save_buffer
                    save_buffer = [] 
                    print("Recode a piece of  voice successfully!")
                    return True
            if time_count==0: 
                if len(save_buffer)>0:
                    self.Voice_String = save_buffer
                    save_buffer = [] 
                    print("Recode a piece of  voice successfully!")
                    return True
                else:
                    return False

if __name__ == "__main__":
    r = recoder()
    r.recoder()
    r.savewav("test.wav")