FCMBP聚类法在语音识别和模糊控制中的应用

FCMBP clustering methods and its applications in speech recognition and fuzzy control

首先从理论上证明了 FCMBP模糊聚类方法 ,即基于摄动的模糊聚类方法 ,比传递闭包法失真小 ;其次 ,用例子说明这两种聚类方法的聚类结果并不总是相同 ,有时还会产生本质差异 ;再次 ,提出用 FCMBP模糊聚类方法设立语音模式的参考向量集来进行语音识别 ,该方法能提高语音识别的正确率 ;最后 ,利用 FCMBP模糊聚类方法滤除量测数据对中的噪声 ,再用这些处理过的数据构造模糊控制规则 ,从而达到模糊控制规则的优化 .

First, it is proved that the FCMBP fuzzy clustering methods, i.e. fuzzy clustering methods based on perturbation, have smaller error than transitive closure methods. Second, it is first pointed that their clustering results are not always the same by example. Moreover, there are differences between them in nature. Third, the FCMBP fuzzy clustering methods are used to speech recognition. The methods can improve the correct rate of speech recognition. Finally, the FCMBP fuzzy clustering methods are applied to filtering the noise in datum. After filtering the noise, the datum is using to construct fuzzy control rules. So the fuzzy control rules are optimized.