基于神经网络的语音识别与自动检验研究

语音识别系统在语音识别中自我判定识别结果，并从错误中自动获取经验改正错误实现知识的自我完善具有重要意义。采用人工神经网络可以有效学习与更新知识，人工神经网络与语音识别结果自动检验方法结合实现一种新的有效学习与更新系统。在该系统中采用基于LEA判别法的梯度牛顿有效结合神经网络快速学习方法。该系统实现在语音识别实践中能够自学习并提高识别率，具有一定的智能。文中给出系统原理图和实验结果。

Study of Speech Recognition and Automation Verifiable Method by Neural Networks

It is very important for the speech recognition system to judge result solely and capture experiences from wrong and overcome mistakes in order to realize automation adjusting of knowledges. By using artificial neural networks, it is effective to learn and refresh knowledges. A new type of effective study and refresh system has been made through combining artificial neural networks with automation verifiable method of speech recognition. The LEA decision method has been given so that grads methods and Newton methods can be effectively combined in neural networks learning. This system can self-study and rise correct rate. We have also given the system principle diagram and the experiment results.