修复了识别成汉字时有时候识别不全的bug

LanguageModel.py

@@ -50,21 +50,32 @@ def SpeechToText(self, list_syllable):
 		if(length == 0): # 传入的参数没有包含任何拼音时
 			return ''
 
+		# 先取出一个字，即拼音列表中第一个字
 		str_tmp = [list_syllable[0]]
+
 		for i in range(0, length - 1):
+			# 依次从第一个字开始每次连续取两个字拼音
 			str_split = list_syllable[i] + ' ' + list_syllable[i+1]
 			#print(str_split,str_tmp,r)
+			# 如果这个拼音在汉语拼音状态转移字典里的话
 			if(str_split in self.pinyin):
+				# 将第二个字的拼音加入
 				str_tmp.append(list_syllable[i+1])
 			else:
+				# 否则不加入，然后直接将现有的拼音序列进行解码
 				str_decode = self.decode(str_tmp, 0.0000)
 				#print('decode ',str_tmp,str_decode)
 				if(str_decode != []):
 					r += str_decode[0][0]
+				# 再重新从i+1开始作为第一个拼音
 				str_tmp = [list_syllable[i+1]]
+
 
-		#print(str_tmp)
+		#print('最后：', str_tmp)
 		str_decode = self.decode(str_tmp, 0.0000)
+
+		#print('剩余解码：',str_decode)
+
 		if(str_decode != []):
 			r += str_decode[0][0]
 
@@ -208,7 +219,7 @@ def GetPinyin(self, filename):
 
 			list_pinyin=pinyin_split[0]
 
-			if(list_pinyin not in dic):
+			if(list_pinyin not in dic and int(pinyin_split[1]) > 1):
 				dic[list_pinyin] = pinyin_split[1]
 		return dic
 

SpeechModel24.py

@@ -396,9 +396,9 @@ def model(self):
 
 	ms = ModelSpeech(datapath)
 
-	ms.LoadModel(modelpath + 'm24\\speech_model24_e_0_step_112500.model')
+	ms.LoadModel(modelpath + 'm24\\speech_model24_e_0_step_411000.model')
 	#ms.TrainModel(datapath, epoch = 50, batch_size = 4, save_step = 500)
-	ms.TestModel(datapath, str_dataset='test', data_count = 128, out_report = True)
+	#ms.TestModel(datapath, str_dataset='test', data_count = 128, out_report = True)
 	#r = ms.RecognizeSpeech_FromFile('E:\\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('E:\\语音数据集\\wav\\train\\A11\\A11_167.WAV')

SpeechModel25.py

@@ -119,7 +119,7 @@ def CreateModel(self):
 
 		model = Model(inputs=[input_data, labels, input_length, label_length], outputs=loss_out)
 
-		model.summary()
+		#model.summary()
 
 		# clipnorm seems to speeds up convergence
 		#sgd = SGD(lr=0.0001, decay=1e-6, momentum=0.9, nesterov=True, clipnorm=5)
@@ -403,8 +403,9 @@ def model(self):
 
 	ms = ModelSpeech(datapath)
 
-	#ms.LoadModel(modelpath + 'm25\\speech_model25_e_0_step_1.model')
-	ms.TrainModel(datapath, epoch = 50, batch_size = 4, save_step = 500)
+	# 全场最佳 speech_model25_e_0_step_340000
+	ms.LoadModel(modelpath + 'm25\\speech_model25_e_0_step_382000.model')
+	#ms.TrainModel(datapath, epoch = 50, batch_size = 4, save_step = 500)
 	#ms.TestModel(datapath, str_dataset='test', data_count = 128, out_report = True)
 	#r = ms.RecognizeSpeech_FromFile('E:\\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')