使用Python和Pocketsphinx进行实时识别
我最近一直在使用python中的pocket sphinx。我成功地得到了那份工作 下面的示例用于识别记录的wav使用Python和Pocketsphinx进行实时识别,python,cmusphinx,Python,Cmusphinx,我最近一直在使用python中的pocket sphinx。我成功地得到了那份工作 下面的示例用于识别记录的wav #!/usr/bin/env python import sys,os def decodeSpeech(hmmd,lmdir,dictp,wavfile): """ Decodes a speech file """ try: import pocketsphinx as ps import sp
#!/usr/bin/env python
import sys,os
def decodeSpeech(hmmd,lmdir,dictp,wavfile):
"""
Decodes a speech file
"""
try:
import pocketsphinx as ps
import sphinxbase
except:
print """Pocket sphinx and sphixbase is not installed
in your system. Please install it with package manager.
"""
speechRec = ps.Decoder(hmm = hmmd, lm = lmdir, dict = dictp)
wavFile = file(wavfile,'rb')
wavFile.seek(44)
speechRec.decode_raw(wavFile)
result = speechRec.get_hyp()
return result[0]
if __name__ == "__main__":
hmdir = "/home/jaganadhg/Desktop/Docs_New/kgisl/model/hmm/wsj1"
lmd = "/home/jaganadhg/Desktop/Docs_New/kgisl/model/lm/wsj/wlist5o.3e-7.vp.tg.lm.DMP"
dictd = "/home/jaganadhg/Desktop/Docs_New/kgisl/model/lm/wsj/wlist5o.dic"
wavfile = "/home/jaganadhg/Desktop/Docs_New/kgisl/sa1.wav"
recognised = decodeSpeech(hmdir,lmd,dictd,wavfile)
print "%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%"
print recognised
print "%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%"
可以调用函数？用于实时识别的代码如下所示：
您还可以在pocketsphinx中使用gstreamer python绑定，请选中并重试。Pocketsphinx现在是一个GStreamer插件。这是我在互联网上看到的代码，我修改了一些东西，以真正听到单词“非常糟糕和缓慢” 你可以帮我永久地修改它。它是基于ubuntu 16.04 LTS构建的 我对编程知之甚少 期待帮助
# -*- encoding: utf-8 -*-
#!/usr/bin/env python
from pocketsphinx.pocketsphinx import *
from sphinxbase.sphinxbase import *
import os
import pyaudio
import wave
import audioop
from collections import deque
import time
import math;import Mic
"""
Written by Sophie Li, 2016
http://blog.justsophie.com/python-speech-to-text-with-pocketsphinx/
"""
class SpeechDetector:
def __init__(self):
# Microphone stream config.
self.CHUNK = 1024 # CHUNKS of bytes to read each time from mic
self.FORMAT = pyaudio.paInt16
self.CHANNELS = 1
self.RATE = 16000
self.SILENCE_LIMIT = 1 # Silence limit in seconds. The max ammount of seconds where
# only silence is recorded. When this time passes the
# recording finishes and the file is decoded
self.PREV_AUDIO = 0.5 # Previous audio (in seconds) to prepend. When noise
# is detected, how much of previously recorded audio is
# prepended. This helps to prevent chopping the beginning
# of the phrase.
self.THRESHOLD = 4500
self.num_phrases = -1
# These will need to be modified according to where the pocketsphinx folder is
MODELDIR = "/home/l/Desktop/pocketsphinx/model/en-us"
# Create a decoder with certain model
config = Decoder.default_config()
config.set_string('-hmm', os.path.join(MODELDIR, '/home/l/Desktop/pocketsphinx/model/en-us/en-us/'))
config.set_string('-lm', os.path.join(MODELDIR, '/home/l/Desktop/pocketsphinx/model/en-us/en-us.lm.bin'))
config.set_string('-dict', os.path.join(MODELDIR, '/home/l/Desktop/pocketsphinx/model/en-us/cmudict-en-us.dict'))
config.set_string('-keyphrase', 'no one')
config.set_float('-kws_threshold', 1e+20)
# Creaders decoder object for streaming data.
self.decoder = Decoder(config)
def setup_mic(self, num_samples=50):
""" Gets average audio intensity of your mic sound. You can use it to get
average intensities while you're talking and/or silent. The average
is the avg of the .2 of the largest intensities recorded.
"""
#print "Getting intensity values from mic."
p = pyaudio.PyAudio()
stream = p.open(format=self.FORMAT,
channels=self.CHANNELS,
rate=self.RATE,
input=True,
frames_per_buffer=self.CHUNK)
values = [math.sqrt(abs(audioop.avg(stream.read(self.CHUNK), 4)))
for x in range(num_samples)]
values = sorted(values, reverse=True)
r = sum(values[:int(num_samples * 0.2)]) / int(num_samples * 0.2)
#print " Finished "
#print " Average audio intensity is ", r
stream.close()
p.terminate()
if r < 3000:
self.THRESHOLD = 3500
else:
self.THRESHOLD = r + 100
def save_speech(self, data, p):
"""
Saves mic data to temporary WAV file. Returns filename of saved
file
"""
filename = 'output_'+str(int(time.time()))
# writes data to WAV file
data = ''.join(data)
wf = wave.open(filename + '.wav', 'wb')
wf.setnchannels(1)
wf.setsampwidth(p.get_sample_size(pyaudio.paInt16))
wf.setframerate(16000) # TODO make this value a function parameter?
wf.writeframes(data)
wf.close()
return filename + '.wav'
def decode_phrase(self, wav_file):
self.decoder.start_utt()
stream = open(wav_file, "rb")
while True:
buf = stream.read(1024)
if buf:
self.decoder.process_raw(buf, False, False)
else:
break
self.decoder.end_utt()
words = []
[words.append(seg.word) for seg in self.decoder.seg()]
return words
def run(self):
"""
Listens to Microphone, extracts phrases from it and calls pocketsphinx
to decode the sound
"""
self.setup_mic()
#Open stream
p = pyaudio.PyAudio()
stream = p.open(format=self.FORMAT,
channels=self.CHANNELS,
rate=self.RATE,
input=True,
frames_per_buffer=self.CHUNK)
audio2send = []
cur_data = '' # current chunk of audio data
rel = self.RATE/self.CHUNK
slid_win = deque(maxlen=self.SILENCE_LIMIT * rel)
#Prepend audio from 0.5 seconds before noise was detected
prev_audio = deque(maxlen=self.PREV_AUDIO * rel)
started = False
while True:
cur_data = stream.read(self.CHUNK)
slid_win.append(math.sqrt(abs(audioop.avg(cur_data, 4))))
if sum([x > self.THRESHOLD for x in slid_win]) > 0:
if started == False:
print "Bắt đầu ghi âm"
started = True
audio2send.append(cur_data)
elif started:
print "Hoàn thành ghi âm"
filename = self.save_speech(list(prev_audio) + audio2send, p)
r = self.decode_phrase(filename)
print "RESULT: ", r
# hot word for me " no one" if r.count('one') and r.count("no") > 0 the end programs
if r.count("one") > 0 and r.count("no") > 0:
Mic.playaudiofromAudio().play("/home/l/Desktop/PROJECT/Audio/beep_hi.wav")
os.remove(filename)
return
# Removes temp audio file
os.remove(filename)
# Reset all
started = False
slid_win = deque(maxlen=self.SILENCE_LIMIT * rel)
prev_audio = deque(maxlen= 0.5 * rel)
audio2send = []
print "Chế độ nghe ..."
else:
prev_audio.append(cur_data)
print "* Hoàn thành nghe"
stream.close()
p.terminate()
#-*-编码：utf-8-*-
#!/usr/bin/env python
从pocketsphinx.pocketsphinx导入*
从sphinxbase.sphinxbase导入*
导入操作系统
导入pyaudio
输入波
导入音频操作
从集合导入deque
导入时间
导入数学；进口话筒
"""
作者：Sophie Li，2016
http://blog.justsophie.com/python-speech-to-text-with-pocketsphinx/
"""
课堂演讲检测器：
定义初始化（自）：
#麦克风流配置。
self.CHUNK=1024#每次从麦克风读取的字节块
self.FORMAT=pyaudio.paInt16
self.CHANNELS=1
自费率=16000
self.SILENCE_LIMIT=1#以秒为单位的SILENCE LIMIT。最大秒数，其中
#只有沉默被记录下来。当这段时间过去的时候
#录制完成，文件被解码
self.PREV_AUDIO=0.5#前一个音频（以秒为单位）前置。当噪音
#检测到之前录制的音频有多少
#预先准备好的。这有助于防止切碎开始
#这句话的意思。
self.THRESHOLD=4500
self.num_短语=-1
#这些将需要根据pocketsphinx文件夹所在的位置进行修改
MODELDIR=“/home/l/Desktop/pocketsphinx/model/en-us”
#创建具有特定模型的解码器
config=Decoder.default\u config（）
config.set_字符串（'-hmm'，os.path.join（MODELDIR，'/home/l/Desktop/pocketsphinx/model/en-us/en-us/'））
config.set_字符串（'-lm'，os.path.join（MODELDIR，'/home/l/Desktop/pocketsphinx/model/en-us/en-us.lm.bin'））
config.set_字符串（'-dict'，os.path.join（MODELDIR，'/home/l/Desktop/pocketsphinx/model/en-us/cmudit en-us.dict'））
config.set_字符串（'-keyphase'，noone'）
配置设置浮点数（'-kws_阈值'，1e+20）
#用于流数据的Creaders解码器对象。
self.decoder=解码器（配置）
def设置麦克风（自身，样本数=50）：
“”获取麦克风声音的平均音频强度。您可以使用它来获取
你说话和/或沉默时的平均强度
是记录的最大强度中.2的平均值。
"""
#打印“从麦克风获取强度值”
p=pyaudio.pyaudio（）
流=p.open（格式=self.format，
通道=自通道，
速率=自速率，
输入=真，
帧（每个缓冲区=self.CHUNK）
value=[math.sqrt（abs（audioop.avg（stream.read（self.CHUNK），4）））
对于范围内的x（数量样本）]
值=已排序（值，反向=真）
r=总和（值[：int（num_samples*0.2）]）/int（num_samples*0.2）
#打印“已完成”
#打印“平均音频强度为”，r
stream.close（）
p、 终止（）
如果r<3000：
自阈值=3500
其他：
自阈值=r+100
def save_语音（自我、数据、p）：
"""
将麦克风数据保存到临时WAV文件。返回保存的麦克风数据的文件名
文件
"""
文件名='output\'+str（int（time.time（）））
#将数据写入WAV文件
数据=“”。连接（数据）
wf=wave.open（文件名+'.wav'，wb'）
wf.设置通道（1）
wf.setsampwidth（p.get_样本大小（pyaudio.paInt16））
setframerate（16000）#如何将此值设为函数参数？
wf.writeframes（数据）
wf.close（）
返回文件名+'.wav'
def解码_短语（自身、wav_文件）：
self.decoder.start_utt（）
流=打开（wav_文件，“rb”）
尽管如此：
buf=流读取（1024）
如果buf：
self.decoder.process_raw（buf，False，False）
其他：
打破
self.decoder.end_utt（）
单词=[]
[self.decoder.seg（）中seg的words.append（seg.word）]
回话
def运行（自）：
"""
听麦克风，从中提取短语并呼叫Pocketsphenx
解码声音
"""
self.setup_mic（）
#明流
p=pyaudio.pyaudio（）
流=p.open（格式=self.format，
通道=自通道，
速率=自速率，
输入=真，
帧（每个缓冲区=self.CHUNK）
audio2send=[]
cur_data=''当前音频数据块
rel=self.RATE/self.CHUNK
滑动\u win=deque（maxlen=self.silen\u LIMIT*rel）
#在检测到噪音前0.5秒开始预编音频
prev_audio=deque（maxlen=self.prev_audio
# -*- encoding: utf-8 -*-
#!/usr/bin/env python
from pocketsphinx.pocketsphinx import *
from sphinxbase.sphinxbase import *
import os
import pyaudio
import wave
import audioop
from collections import deque
import time
import math;import Mic
"""
Written by Sophie Li, 2016
http://blog.justsophie.com/python-speech-to-text-with-pocketsphinx/
"""
class SpeechDetector:
def __init__(self):
# Microphone stream config.
self.CHUNK = 1024 # CHUNKS of bytes to read each time from mic
self.FORMAT = pyaudio.paInt16
self.CHANNELS = 1
self.RATE = 16000
self.SILENCE_LIMIT = 1 # Silence limit in seconds. The max ammount of seconds where
# only silence is recorded. When this time passes the
# recording finishes and the file is decoded
self.PREV_AUDIO = 0.5 # Previous audio (in seconds) to prepend. When noise
# is detected, how much of previously recorded audio is
# prepended. This helps to prevent chopping the beginning
# of the phrase.
self.THRESHOLD = 4500
self.num_phrases = -1
# These will need to be modified according to where the pocketsphinx folder is
MODELDIR = "/home/l/Desktop/pocketsphinx/model/en-us"
# Create a decoder with certain model
config = Decoder.default_config()
config.set_string('-hmm', os.path.join(MODELDIR, '/home/l/Desktop/pocketsphinx/model/en-us/en-us/'))
config.set_string('-lm', os.path.join(MODELDIR, '/home/l/Desktop/pocketsphinx/model/en-us/en-us.lm.bin'))
config.set_string('-dict', os.path.join(MODELDIR, '/home/l/Desktop/pocketsphinx/model/en-us/cmudict-en-us.dict'))
config.set_string('-keyphrase', 'no one')
config.set_float('-kws_threshold', 1e+20)
# Creaders decoder object for streaming data.
self.decoder = Decoder(config)
def setup_mic(self, num_samples=50):
""" Gets average audio intensity of your mic sound. You can use it to get
average intensities while you're talking and/or silent. The average
is the avg of the .2 of the largest intensities recorded.
"""
#print "Getting intensity values from mic."
p = pyaudio.PyAudio()
stream = p.open(format=self.FORMAT,
channels=self.CHANNELS,
rate=self.RATE,
input=True,
frames_per_buffer=self.CHUNK)
values = [math.sqrt(abs(audioop.avg(stream.read(self.CHUNK), 4)))
for x in range(num_samples)]
values = sorted(values, reverse=True)
r = sum(values[:int(num_samples * 0.2)]) / int(num_samples * 0.2)
#print " Finished "
#print " Average audio intensity is ", r
stream.close()
p.terminate()
if r < 3000:
self.THRESHOLD = 3500
else:
self.THRESHOLD = r + 100
def save_speech(self, data, p):
"""
Saves mic data to temporary WAV file. Returns filename of saved
file
"""
filename = 'output_'+str(int(time.time()))
# writes data to WAV file
data = ''.join(data)
wf = wave.open(filename + '.wav', 'wb')
wf.setnchannels(1)
wf.setsampwidth(p.get_sample_size(pyaudio.paInt16))
wf.setframerate(16000) # TODO make this value a function parameter?
wf.writeframes(data)
wf.close()
return filename + '.wav'
def decode_phrase(self, wav_file):
self.decoder.start_utt()
stream = open(wav_file, "rb")
while True:
buf = stream.read(1024)
if buf:
self.decoder.process_raw(buf, False, False)
else:
break
self.decoder.end_utt()
words = []
[words.append(seg.word) for seg in self.decoder.seg()]
return words
def run(self):
"""
Listens to Microphone, extracts phrases from it and calls pocketsphinx
to decode the sound
"""
self.setup_mic()
#Open stream
p = pyaudio.PyAudio()
stream = p.open(format=self.FORMAT,
channels=self.CHANNELS,
rate=self.RATE,
input=True,
frames_per_buffer=self.CHUNK)
audio2send = []
cur_data = '' # current chunk of audio data
rel = self.RATE/self.CHUNK
slid_win = deque(maxlen=self.SILENCE_LIMIT * rel)
#Prepend audio from 0.5 seconds before noise was detected
prev_audio = deque(maxlen=self.PREV_AUDIO * rel)
started = False
while True:
cur_data = stream.read(self.CHUNK)
slid_win.append(math.sqrt(abs(audioop.avg(cur_data, 4))))
if sum([x > self.THRESHOLD for x in slid_win]) > 0:
if started == False:
print "Bắt đầu ghi âm"
started = True
audio2send.append(cur_data)
elif started:
print "Hoàn thành ghi âm"
filename = self.save_speech(list(prev_audio) + audio2send, p)
r = self.decode_phrase(filename)
print "RESULT: ", r
# hot word for me " no one" if r.count('one') and r.count("no") > 0 the end programs
if r.count("one") > 0 and r.count("no") > 0:
Mic.playaudiofromAudio().play("/home/l/Desktop/PROJECT/Audio/beep_hi.wav")
os.remove(filename)
return
# Removes temp audio file
os.remove(filename)
# Reset all
started = False
slid_win = deque(maxlen=self.SILENCE_LIMIT * rel)
prev_audio = deque(maxlen= 0.5 * rel)
audio2send = []
print "Chế độ nghe ..."
else:
prev_audio.append(cur_data)
print "* Hoàn thành nghe"
stream.close()
p.terminate()