Python调用百度语音识别REST API

nxhujiee 2016-03-27

展开全文

百度语音识别通过 REST API 的方式给开发者提供一个通用的 HTTP 接口，基于该接口，开发者可以轻松的获取语音识别能力。SDK中只提供了PHP、C和JAVA的相关样例，然而个人以为，使用Python开发难度更低，本文描述了简单使用Python调用百度语音识别服务 REST API 的简单样例。

注册开发者账号和创建应用不再赘述，百度的REST API在调用过程基本分为三步：

获取token
提交数据
处理JSON

获取token

#Api_Key申请的api_key
#Secrect_Key对应的secret_key
url = 'https://openapi.baidu.com/oauth/2.0/token?grant_type=client_credentials&client_id='+Api_Key+'&client_secret='+Secrect_Key
res = urllib2.urlopen(url).read()
data = json.loads(res)
token = data['access_token']    #获取的token
print 'token获取成功：'+`token`

提交数据

VOICE_RATE = 8000
WAVE_FILE = 'temp.wav'
USER_ID = 'duvoice'
WAVE_TYPE = 'wav'
#其它参数可参考sdk文档

f = open(WAVE_FILE,'r')
speech = base64.b64encode(f.read())
size = os.path.getsize(WAVE_FILE)
update = json.dumps({'format':WAVE_TYPE,'rate':VOICE_RATE,'channel':1,'cuid':USER_ID,'token':token,'speech':speech,'len':size})
r = urllib2.urlopen(self.url,update)

处理JSON

t = r.read()
result = json.loads(t)
if result['err_msg']=='success.':
    word = result['result'][0].encode('utf-8')
    if word!='':
        if word[len(word)-3:len(word)]=='，':
            print word[0:len(word)-3]
            return word[0:len(word)-3]
        else:
            print word
            return word
    else:
        print "音频文件不存在或格式错误"
        return ''
else:
    print "错误"

经过以上步骤便可提交一个wav文件进行解析，同时也可使用python的pyaudio库进行录音：

#!/usr/bin/env python
# -*- coding: utf-8 -*-

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

class recoder:
    NUM_SAMPLES = 2000      #pyaudio内置缓冲大小
    SAMPLING_RATE = 8000    #取样频率
    LEVEL = 500         #声音保存的阈值
    COUNT_NUM = 20      #NUM_SAMPLES个取样之内出现COUNT_NUM个大于LEVEL的取样则记录声音
    SAVE_LENGTH = 8         #声音记录的最小长度：SAVE_LENGTH * NUM_SAMPLES 个取样
    TIME_COUNT = 60     #录音时间，单位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("".join(self.Voice_String)) 
        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
            # 读入NUM_SAMPLES个取样
            string_audio_data = stream.read(self.NUM_SAMPLES) 
            # 将读入的数据转换为数组
            audio_data = np.fromstring(string_audio_data, dtype=np.short)
            # 计算大于LEVEL的取样的个数
            large_sample_count = np.sum( audio_data > self.LEVEL )
            print np.max(audio_data)
            # 如果个数大于COUNT_NUM，则至少保存SAVE_LENGTH个块
            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 : 
            # 将要保存的数据存放到save_buffer中
                #print  save_count > 0 and time_count >0
                save_buffer.append( string_audio_data ) 
            else: 
            #print save_buffer
            # 将save_buffer中的数据写入WAV文件，WAV文件的文件名是保存的时刻
                #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