def asr_mandarin(req_file: "UploadedFile"):
load_data(req_file)
keras.backend.clear_session()
ms = ModelSpeech(asrPath)
ms.LoadModel(modelpath + 'speech_model251_e_0_step_625000.model')
r = ms.RecognizeSpeech_FromFile(filePath)
ml = ModelLanguage('model_language')
ml.LoadModel()
str_pinyin = r
r = ml.SpeechToText(str_pinyin)
return r
def speech_recognition(f):
datapath = '.'
modelpath = 'model_speech' + '\\'
ms = ModelSpeech(datapath)
ms.LoadModel(modelpath + 'speech_model251_e_0_step_625000.model')
r = ms.RecognizeSpeech_FromFile(f)
K.clear_session()
print('*[提示] 语音识别结果：\n', r)
ml = ModelLanguage('model_language')
ml.LoadModel()
str_pinyin = r
r = ml.SpeechToText(str_pinyin)
print('语音转文字结果：\n', r)
return r
def predict(video_file):
from SpeechModel251 import ModelSpeech
from LanguageModel2 import ModelLanguage
from keras import backend as K
datapath = 'dataset'
modelpath = 'model_speech'
ms = ModelSpeech(datapath)
ms.LoadModel(modelpath + '/m251/speech_model251_e_0_step_60500.model')
pinyin = ms.RecognizeSpeech_FromFile(video_file)
K.clear_session()
ml = ModelLanguage('model_language')
ml.LoadModel()
str_pinyin = pinyin
text = ml.SpeechToText(str_pinyin)
return pinyin, text
def CTC_tf(current_path):
datapath = ''
modelpath = 'model_speech'
system_type = plat.system() # 由于不同的系统的文件路径表示不一样，需要进行判断
if (system_type == 'Windows'):
datapath = current_path
modelpath = modelpath + '\\'
elif (system_type == 'Linux'):
datapath = 'dataset'
modelpath = modelpath + '/'
else:
print('*[Message] Unknown System\n')
datapath = 'dataset'
modelpath = modelpath + '/'
ms = ModelSpeech(datapath)
ms.LoadModel(modelpath + 'speech_model251_e_0_step_12000.model')
#ms.TestModel(datapath, str_dataset='test', data_count = 64, out_report = True)
rr = ms.RecognizeSpeech_FromFile(current_path + '\\chunk-00.wav')
print('*[提示] 语音识别结果：\n', rr)
ml = ModelLanguage('model_language')
ml.LoadModel()
#str_pinyin = ['zhe4','zhen1','shi4','ji2', 'hao3','de5']
#str_pinyin = ['jin1', 'tian1', 'shi4', 'xing1', 'qi1', 'san1']
#str_pinyin = ['ni3', 'hao3','a1']
str_pinyin = rr
#str_pinyin = ['su1', 'bei3', 'jun1', 'de5', 'yi4','xie1', 'ai4', 'guo2', 'jiang4', 'shi4', 'ma3', 'zhan4', 'shan1', 'ming2', 'yi1', 'dong4', 'ta1', 'ju4', 'su1', 'bi3', 'ai4', 'dan4', 'tian2','mei2', 'bai3', 'ye3', 'fei1', 'qi3', 'kan4', 'zhan4']
r = ml.SpeechToText(str_pinyin)
print('语音转文字结果：\n', r)
ctc_result = hanzi_pinyin(r)
return ctc_result
datapath = 'C:\\test'
modelpath = modelpath + '\\'
elif (system_type == 'Linux'):
datapath = 'dataset'
modelpath = modelpath + '/'
else:
print('*[Message] Unknown System\n')
datapath = 'dataset'
modelpath = modelpath + '/'
ms = ModelSpeech(datapath)
ms.LoadModel(modelpath + 'speech_model251_e_0_step_12000.model')
#ms.TestModel(datapath, str_dataset='test', data_count = 64, out_report = True)
r = ms.RecognizeSpeech_FromFile('C:\\test\\online_star.wav')
print('*[提示] 语音识别结果：\n', r)
ml = ModelLanguage('model_language')
ml.LoadModel()
#str_pinyin = ['zhe4','zhen1','shi4','ji2', 'hao3','de5']
#str_pinyin = ['jin1', 'tian1', 'shi4', 'xing1', 'qi1', 'san1']
#str_pinyin = ['ni3', 'hao3','a1']
str_pinyin = r
#str_pinyin = ['su1', 'bei3', 'jun1', 'de5', 'yi4','xie1', 'ai4', 'guo2', 'jiang4', 'shi4', 'ma3', 'zhan4', 'shan1', 'ming2', 'yi1', 'dong4', 'ta1', 'ju4', 'su1', 'bi3', 'ai4', 'dan4', 'tian2','mei2', 'bai3', 'ye3', 'fei1', 'qi3', 'kan4', 'zhan4']
r = ml.SpeechToText(str_pinyin)
print('语音转文字结果：\n', r)
class App(QWidget):
def __init__(self):
super(App, self).__init__()
self.record_name = 'saved_record.wav'
self.ms = ModelSpeech('dataset')
self.ms.LoadModel('model_speech/speech_model251_e_0_step_12000.model')
self.ml = ModelLanguage('model_language')
self.ml.LoadModel()
self.title = 'ASR demo'
self.left = 10
self.top = 10
self.width = 420
self.height = 400
self.rec = Recorder()
#self.rec.start_thread()
self.initUI()
self.rec.start()
#self.setWindowTitle("ASR demo")
def initUI(self):
self.setWindowTitle(self.title)
self.setGeometry(self.left, self.top, self.width, self.height)
button = QPushButton('Record', self)
button.setToolTip('Press to start recording')
button.move(100, 70)
button.clicked.connect(self.start_record)
button = QPushButton('To Transcript', self)
button.setToolTip('Press to convert to transcript')
button.move(200, 70)
button.clicked.connect(self.stop_record)
button = QPushButton('Clear', self)
button.setToolTip('Press to clear transcripts')
button.move(100, 100)
button.clicked.connect(self.clear)
self.text_edit = QTextEdit("What you said: ", self)
self.text_edit.setReadOnly(True)
self.text_edit.move(100, 140)
#self.results=QLabel(self)
#self.results.move(100,140)
self.show()
@pyqtSlot()
def clear(self):
self.text_edit.clear()
self.text_edit.append("What you said: ")
@pyqtSlot()
def start_record(self):
self.rec.start()
#print('PyQt5 button click')
#self.rec.start()
@pyqtSlot()
def stop_record(self):
print(len(self.rec._frames))
#print('PyQt5 button click')
self.rec.stop()
self.rec.save(self.record_name)
r = self.ms.RecognizeSpeech_FromFile(self.record_name)
self.w = self.ml.SpeechToText(r)
print('语音转文字结果：\n', self.w)
self.text_edit.append(self.w)