Python神技之利用百度AI声控电脑关机

这次我们来看看如何利用百度AI来声控电脑关机。首先需要安装百度AI的Python SDK，并且创建语音识别的应用，获取AppID、API Key、Secret Key这三项内容，以便在我们写的程序里使用，详情可见上上篇文章，这里就不赘述了。

完整代码如下：

1# coding: utf-8

2"""

3@author: Kevin Wong

4@function: python实现的录音及语音识别程序

5@time: 2018/11/15 23:14

6"""

7import os

8import wave

9import numpy as np

10from pyaudio import PyAudio, paInt16

11from aip import AipSpeech

12import os

13

14

15class Recorder(object):

16 def __init__(self):

17 # pyaudio内置缓冲大小

18 self.num_samples = 2000

19 # 取样频率

20 self.sampling_rate = 8000

21 # 声音保存的阈值

22 self.level = 1500

23 # count_num个取样之内出现COUNT_NUM个大于LEVEL的取样则记录声音

24 self.count_num = 20

25 # 声音记录的最小长度：save_length * num_samples 个取样

26 self.save_length = 8

27 # 录音时间，单位s

28 self.time_count = 8

29 self.voice_string = []

30

31 # 保存为音频文件

32 def save_wav(self, filename):

33 wf = wave.open(filename, 'wb')

34 wf.setnchannels(1)

35 wf.setsampwidth(2)

36 wf.setframerate(self.sampling_rate)

37 wf.writeframes(np.array(self.voice_string).tostring())

38 wf.close()

39

40 # 读取音频

41 def recorder(self):

42 pa = PyAudio()

43 stream = pa.open(format=paInt16, channels = 1, rate = self.sampling_rate, input = True, frames_per_buffer = self.num_samples)

44 save_count = 0

45 save_buffer = []

46 time_count = self.time_count

47 while True:

48 time_count -= 1

49 # 读入num_samples个取样

50 string_audio_data = stream.read(self.num_samples)

51 # 将读入的数据转换为数组

52 audio_data = np.fromstring(string_audio_data, dtype = np.short)

53 # 计算大于 level 的取样的个数

54 large_sample_count = np.sum(audio_data > self.level)

55 print(np.max(audio_data)), "large_sample_count=>", large_sample_count

56 # 如果个数大于COUNT_NUM，则至少保存SAVE_LENGTH个块

57 if large_sample_count > self.count_num:

58 save_count = self.save_length

59 else:

60 save_count -=1

61 if save_count

62 save_count = 0

63 if save_count > 0:

64 save_buffer.append(string_audio_data)

65 else:

66 if len(save_buffer) > 0:

67 self.voice_string = save_buffer

68 save_buffer = []

69 print("Recode a piece of voice successfully!")

70 return True

71 if time_count == 0:

72 if len(save_buffer) > 0:

73 self.voice_string = save_buffer

74 save_buffer = []

75 print("Recode a piece of voice successfully!")

76 return True

77 else:

78 return False

79 return True

80

81

82# 读取本地音频文件

83def get_file_content(filePath):

84 with open(filePath, 'rb') as fp:

85 return fp.read()

86

87if __name__ == '__main__':

88 """ 你的 APPID AK SK """

89 APP_ID = '14810929'

90 API_KEY = 'hD1sGacRqCWybF9lBqumMriS'

91 SECRET_KEY = 'zKtG8uv3mv4tKqC5avL1ua9YGM38YAAG'

92 client = AipSpeech(APP_ID, API_KEY, SECRET_KEY)

93

94 while True:

95 recorder = Recorder()

96 # 开始录音

97 recorder.recorder()

98 # 保存音频文件

99 recorder.save_wav("01.wav")

100 # 识别本地文件

101 res = client.asr(get_file_content('01.wav'), 'wav', 8000, {

102 'dev_pid': 1536,

103 })

104 print(res['result'][0])

105 try:

106 if "计算器" == res['result'][0]:

107 os.system("calc")

108 elif "关机" == res['result'][0]:

109 os.system("shutdown -s -t 300")

110 elif "取消关机" == res['result'][0]:

111 os.system("shutdown -a")

112 elif "退出程序" == res['result'][0]:

113 break

114 except:

115 pass

这里采用了面向对象的编程风格，第15-79行定义了一个Recorder类，其主要功能是对音频文件进行处理，包括将程序运行后将用户的声音以二进制流的形式读取并保存为wav格式的音频文件， 第82行到第85行读取生成的音频文件，并返回文件内容。第87行是主线程入口，只要用户没有对电脑说“退出程序”，就会一直执行while循环读取用户的声音，将音频文件交给百度AI的语音识别接口，并返回识别的文字内容。根据识别的内容，调用Python的os库执行相应的操作。

运行程序后，对电脑说一声“关机”，运行结果如下：

完结，撒花，ye~