SOM结合MLP的神经网络语音识别系统

A New Speech Recognition System Using SOM and MLP Neural Networks

提出一种结合自组织特征映射（Ｓｅｌｆ－ｏｒｇａｎｉｚｉｎｇＦｅａ－ｔｕｒｅＭａｐ，ＳＯＭ）和多层感知器（ＭｕｌｔｉｌａｙｅｒＰｅｒｃｅｐｔｒｏｎ，ＭＬＰ）的神经网络语音识别系统。

A new speech recognition system which combines SOM (Self organizing Feature Map) with MLP (Multilayer Perceptron) neural networks for isolated word recognition is presented. Experimental results show that the recognition performance of the new system is much better than that of the conventional MLP recognizer. The new system is expected to use in more words data base.