基于混合DBNN-BLSTM模型的大词汇量连续语音识别

摘要

The recognition rate is not ideal when the feature extraction is performed on the deep confidence neural network(DBNN)model and the bidirectional long-short term memory (BLSTM),the long-short term memory(LSTM)and BLSTM can better analyze the character-istics of speech data.By combining the DBNN model with BLSTM,a new acoustic modeling method for large vocabulary continuous speech recognition(LVCSR)is proposed and experi-mentally studied based on Keras deep learning framework.The experimental results show that the improved DBNN-BLSTM model has a high recognition accuracy,and the speech recognition rate is 5% higher than that of BLSTM.%深度置信神经网络(DBNN)模型和双向长短时记忆神经网络模型(BLSTM)在单独进行特征提取时识别率不理想,长短时记忆单元(LSTM)与BLSTM模型可以更好解析语音数据特征.因此将DBNN模型和BLSTM 模型相结合,提出一种大词汇量连续语音识别(LVCSR)的声学模型建立方法,并在Keras深度学习框架下进行实验.实验结果表明,使用改进的DBNN-BLSTM模型进行大词汇量连续语音识别,识别精度有所提高,比BLSTM模型的语音识别率提高5%.