一种抗噪孤立字语音识别模型

论提出一种连续隐Markov模型和BP神经网络相结合的，具有两次辨识过程的抗噪孤立字识别模型，首先以连续隐Markov模型完成语音信号的时序建模并提供一次识别信息，以BP神经网络进行后处理，提取二次识别信息，识别结果由两次识别信息共同决定，实验证明，由于有效地利用了隐Markov模型的强时序信号处理能力和BP神经网络的强模式分类和泛化性能，这种识别模型明显地改善了孤立字识别系统的抗噪性能。

A Model for Recognition of Isolated Words in Noisy Environments

To recognize isolated words in noisy environments， A new robust recognition model with dual recognition procedures based on continuo us Hidden Markov Models(CHMM) and BP neural networks(BPNN) is presented. First, the CHMM is applied as the front-end to process the time sequence of speech and the primary recognition information is provided in this s tep. In the next step, BPNN is applied as the back-end and because of its super ior functions of pattern classification and generalization, the primary recognit ion information is non-linearly mapped into the secondary recognition informati on. The final recognition procedure is accomplished with the two kinds of recogn ition information. Experiments prove that using this robust model, recognition r ate can be noticeably improved in noisy environments.