From the project paper:

1     Introduction
In the Speech Technology Project of 2004, a group of students in the Master’s phase of Arti?cial Intelligence were assigned the task of building a speech recogniser for Dutch using Hidden Markov Models (HMM’s). The goal of this project was to build a robust phoneme driven recogniser. That means that it should be able to generalise both from speaker speci?c properties and that its training should be more than just instance based learning. In the HMM paradigm this is supposed to be the case, but we wanted to put this to practice. Also, the system should serve as a stepping stone for future projects.
   ...
    This report is one of the products of the project. The other products are a HMM for Dutch spoken language (albeit for a very restricted domain) and a recipe for building HMM speech recognisers, especially for Dutch and using HTK.

The acoustic models for the Julius speech recognition engine are build using the HTK toolkit.

Ken