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

更新拼音参数和几条文件路径，声学模型文件跟之前版本不再兼容，需要重新训练

This commit is contained in:
nl 2019-03-29 14:28:01 +08:00
parent 07c9b3600b
commit 6579229d7d
6 changed files with 17 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
SpeechModel24.py
View File

@ -28,9 +28,9 @@ class ModelSpeech(): # 语音模型类
def __init__(self, datapath): def __init__(self, datapath):
''' '''
初始化 初始化
默认输出的拼音的表示大小是1422，即1421个拼音+1个空白块 默认输出的拼音的表示大小是1424，即1423个拼音+1个空白块
''' '''
MS_OUTPUT_SIZE = 1422 MS_OUTPUT_SIZE = 1424
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

4
SpeechModel25.py
View File

@ -28,9 +28,9 @@ class ModelSpeech(): # 语音模型类
def __init__(self, datapath): def __init__(self, datapath):
''' '''
初始化 初始化
默认输出的拼音的表示大小是1422，即1421个拼音+1个空白块 默认输出的拼音的表示大小是1424，即1423个拼音+1个空白块
''' '''
MS_OUTPUT_SIZE = 1422 MS_OUTPUT_SIZE = 1424
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

4
SpeechModel251.py
View File

@ -32,9 +32,9 @@ class ModelSpeech(): # 语音模型类
def __init__(self, datapath): def __init__(self, datapath):
''' '''
初始化 初始化
默认输出的拼音的表示大小是1422，即1421个拼音+1个空白块 默认输出的拼音的表示大小是1424，即1423个拼音+1个空白块
''' '''
MS_OUTPUT_SIZE = 1422 MS_OUTPUT_SIZE = 1424
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

4
SpeechModel251_p.py
View File

@ -32,9 +32,9 @@ class ModelSpeech(): # 语音模型类
def __init__(self, datapath): def __init__(self, datapath):
''' '''
初始化 初始化
默认输出的拼音的表示大小是1422，即1421个拼音+1个空白块 默认输出的拼音的表示大小是1424，即1423个拼音+1个空白块
''' '''
MS_OUTPUT_SIZE = 1422 MS_OUTPUT_SIZE = 1424
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

4
SpeechModel26.py
View File

@ -29,9 +29,9 @@ class ModelSpeech(): # 语音模型类
def __init__(self, datapath): def __init__(self, datapath):
''' '''
初始化 初始化
默认输出的拼音的表示大小是1422，即1421个拼音+1个空白块 默认输出的拼音的表示大小是1424，即1423个拼音+1个空白块
''' '''
MS_OUTPUT_SIZE = 1422 MS_OUTPUT_SIZE = 1424
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

14
test.py
View File

@ -8,7 +8,7 @@
import platform as plat import platform as plat
from SpeechModel251 import ModelSpeech from SpeechModel251 import ModelSpeech
from LanguageModel import ModelLanguage from LanguageModel2 import ModelLanguage
from keras import backend as K from keras import backend as K
datapath = '' datapath = ''
@ -16,7 +16,7 @@ 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 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'): if(system_type == 'Windows'):
datapath = 'E:\\语音数据集' datapath = 'D:\\语音数据集'
modelpath = modelpath + '\\' modelpath = modelpath + '\\'
elif(system_type == 'Linux'): elif(system_type == 'Linux'):
datapath = 'dataset' datapath = 'dataset'
@ -29,14 +29,14 @@ else:
ms = ModelSpeech(datapath) ms = ModelSpeech(datapath)
#ms.LoadModel(modelpath + 'm22_2\\0\\speech_model22_e_0_step_257000.model') #ms.LoadModel(modelpath + 'm22_2\\0\\speech_model22_e_0_step_257000.model')
ms.LoadModel(modelpath + 'm251\\speech_model251_e_0_step_117000.model') ms.LoadModel(modelpath + 'm251\\speech_model251_e_0_step_12000.model')
#ms.TestModel(datapath, str_dataset='test', data_count = 64, out_report = True) #ms.TestModel(datapath, str_dataset='test', data_count = 64, out_report = True)
r = ms.RecognizeSpeech_FromFile('D:\\语音数据集\\ST-CMDS-20170001_1-OS\\20170001P00241I0052.wav') r = ms.RecognizeSpeech_FromFile('D:\\语音数据集\\ST-CMDS-20170001_1-OS\\20170001P00241I0052.wav')
#r = ms.RecognizeSpeech_FromFile('E:\语音数据集\ST-CMDS-20170001_1-OS\\20170001P00241I0053.wav') #r = ms.RecognizeSpeech_FromFile('D:\语音数据集\ST-CMDS-20170001_1-OS\\20170001P00241I0053.wav')
#r = ms.RecognizeSpeech_FromFile('E:\\语音数据集\\ST-CMDS-20170001_1-OS\\20170001P00020I0087.wav') #r = ms.RecognizeSpeech_FromFile('D:\\语音数据集\\ST-CMDS-20170001_1-OS\\20170001P00020I0087.wav')
#r = ms.RecognizeSpeech_FromFile('E:\\语音数据集\\wav\\train\\A11\\A11_167.WAV') #r = ms.RecognizeSpeech_FromFile('D:\\语音数据集\\data_thchs30\\data\\A11_167.WAV')
#r = ms.RecognizeSpeech_FromFile('E:\\语音数据集\\wav\\test\\D4\\D4_750.wav') #r = ms.RecognizeSpeech_FromFile('D:\\语音数据集\\data_thchs30\\data\\D4_750.wav')
K.clear_session() K.clear_session()