Checked by the experiments, the improved RBF network has less hidden neural units than before, at the same time keep the accurate of RBF based classifier.
经实验证明,基于改进后的RBF网络具有更少的隐含神经元,但仍然保持了基于RBF网络分类器的准确率。
Here, the HMM is employed to produce a best speech state sequence which is warped to a fixed dimension vector and the RBF neural network is used as classifier.
该方法首先利用HMM生成最佳语音状态序列,然后用函数逼近技术产生对最佳状态序列进行时间规正,最后通过RBF神经网络进行分类识别。
应用推荐