python 语音识别校正_Python语音识别时间偏移量

像这样：from os import environ, path

from pocketsphinx.pocketsphinx import *

from sphinxbase.sphinxbase import *

MODELDIR = "../../../model"

DATADIR = "../../../test/data"

config = Decoder.default_config()

config.set_string('-hmm', path.join(MODELDIR, 'en-us/en-us'))

config.set_string('-lm', path.join(MODELDIR, 'en-us/en-us.lm.bin'))

config.set_string('-dict', path.join(MODELDIR, 'en-us/cmudict-en-us.dict'))

config.set_string('-logfn', '/dev/null')

decoder = Decoder(config)

stream = open(path.join(DATADIR, 'goforward.raw'), 'rb')

in_speech_bf = False

decoder.start_utt()

while True:

buf = stream.read(1024)

if buf:

decoder.process_raw(buf, False, False)

if decoder.get_in_speech() != in_speech_bf:

in_speech_bf = decoder.get_in_speech()

if not in_speech_bf:

decoder.end_utt()

print ('Result:', decoder.hyp().hypstr)

print ([(seg.word, seg.prob, seg.start_frame, seg.end_frame) for seg in decoder.seg()])

decoder.start_utt()

else:

break

decoder.end_utt()

更多示例here。在