The analyzer has to extends org.apache.lucene.analysis.standard.StandardAnalyzer, and overload the method

public TokenStream tokenStream(String fieldName, Reader reader)

to put your own filters. You can have a glance at the example analyzer attached to this article.