OrgLion-ML
/
ASRT_SpeechRecognition
mirror of https://github.com/nl8590687/ASRT_SpeechRecognition.git
1
0
Fork 0
Code Issues Releases Wiki Activity

fix bug

This commit is contained in:
nl8590687 2018-04-28 15:59:32 +08:00
parent 09607561cc
commit e95985d456
3 changed files with 5 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
SpeechModel22.py
View File

@ -34,7 +34,7 @@ class ModelSpeech(): # 语音模型类
初始化 初始化
默认输出的拼音的表示大小是1283，即1282个拼音+1个空白块 默认输出的拼音的表示大小是1283，即1282个拼音+1个空白块
''' '''
MS_OUTPUT_SIZE = 1417 MS_OUTPUT_SIZE = 1419
self.MS_OUTPUT_SIZE = MS_OUTPUT_SIZE # 神经网络最终输出的每一个字符向量维度的大小 self.MS_OUTPUT_SIZE = MS_OUTPUT_SIZE # 神经网络最终输出的每一个字符向量维度的大小
#self.BATCH_SIZE = BATCH_SIZE # 一次训练的batch #self.BATCH_SIZE = BATCH_SIZE # 一次训练的batch
self.label_max_string_length = 64 self.label_max_string_length = 64
@ -87,7 +87,7 @@ class ModelSpeech(): # 语音模型类
#layer_h5 = LSTM(256, activation='relu', use_bias=True, return_sequences=True)(layer_h4) # LSTM层 #layer_h5 = LSTM(256, activation='relu', use_bias=True, return_sequences=True)(layer_h4) # LSTM层
#layer_h6 = Dropout(0.2)(layer_h5) # 随机中断部分神经网络连接，防止过拟合 #layer_h6 = Dropout(0.2)(layer_h5) # 随机中断部分神经网络连接，防止过拟合
layer_h8 = Dense(256, activation="relu", use_bias=True, kernel_initializer='he_normal')(layer_h7) # 全连接层 layer_h8 = Dense(256, activation="relu", use_bias=True, kernel_initializer='he_normal')(layer_h7) # 全连接层
layer_h9 = Dense(1417, use_bias=True, kernel_initializer='he_normal')(layer_h8) # 全连接层 layer_h9 = Dense(self.MS_OUTPUT_SIZE, use_bias=True, kernel_initializer='he_normal')(layer_h8) # 全连接层
y_pred = Activation('softmax', name='Activation0')(layer_h9) y_pred = Activation('softmax', name='Activation0')(layer_h9)
model_data = Model(inputs = input_data, outputs = y_pred) model_data = Model(inputs = input_data, outputs = y_pred)

1
dict.txt
View File

@ -703,6 +703,7 @@ nin2 您
ning2 柠狞凝宁拧咛 ning2 柠狞凝宁拧咛
ning3 拧 ning3 拧
ning4 拧泞宁 ning4 拧泞宁
ni1 妞
niu2 牛 niu2 牛
niu3 扭钮纽 niu3 扭钮纽
nin4 拗 nin4 拗

4
readdata22_2.py
View File

@ -74,9 +74,9 @@ class DataSpeech():
filename_symbollist_thchs30 = 'thchs30' + self.slash + 'train.syllable.txt' filename_symbollist_thchs30 = 'thchs30' + self.slash + 'train.syllable.txt'
filename_symbollist_stcmds = 'st-cmds' + self.slash + 'train.syllable.txt' filename_symbollist_stcmds = 'st-cmds' + self.slash + 'train.syllable.txt'
elif(self.type=='dev'): elif(self.type=='dev'):
filename_wavlist_thchs30 = 'thchs30' + self.slash + 'dev.wav.lst' filename_wavlist_thchs30 = 'thchs30' + self.slash + 'cv.wav.lst'
filename_wavlist_stcmds = 'st-cmds' + self.slash + 'dev.wav.txt' filename_wavlist_stcmds = 'st-cmds' + self.slash + 'dev.wav.txt'
filename_symbollist_thchs30 = 'thchs30' + self.slash + 'dev.syllable.txt' filename_symbollist_thchs30 = 'thchs30' + self.slash + 'cv.syllable.txt'
filename_symbollist_stcmds = 'st-cmds' + self.slash + 'dev.syllable.txt' filename_symbollist_stcmds = 'st-cmds' + self.slash + 'dev.syllable.txt'
elif(self.type=='test'): elif(self.type=='test'):
filename_wavlist_thchs30 = 'thchs30' + self.slash + 'test.wav.lst' filename_wavlist_thchs30 = 'thchs30' + self.slash + 'test.wav.lst'