1 Introduction

This document describes frequently occuring problems and their solutions. Each problem-solution pair is called a “recipe“, which resembles the term “recipe“ used in cooking.

Chapter 2: Learning HARK

 
This chapter include recipes for beginners. These recipes describe methods of recording, localizing, separating and recognizing sound. It is recommended for first-time HARK users to read the recipes found in this chapter.

Chapter 3: Something is Wrong!

 
This chapter describes troubleshooting common problems that occurs during installation and recording. It also includes debugging recipes. This can be used as reference in solving common problems.

Chapter 4: Microphone Array

 
This chapter include recipes on the layout of microphone arrays. Use this chapter as reference when the problem occuring is related to the number of microphones, type of microphones, or in the set-up of microphones in robot systems.

Chapter 5: Input Data Generation

 
This chapter include recipes for generating input data in HARK. Basically, it has recipes for multichannel recording, impulse response measurements, and also multichannel sound generated through simulation.

Chapter 6: Acoustic and Language Models

 
This chapter describes how to build an acoustic model and a language model, which is a requirement in using the speech recognition software Julius supported by HARK.

Chapter 7: HARK-Designer

 
The robot audition system in HARK is created through a network generation GUI ‘HARK-Designer‘ by placing and connecting nodes. This chapter include recipes in using HARK-Designer.

Chapter 8: Sound Source Localization

 
This chapter include recipes for sound source localization. It covers the recipes in building sound source localization system, parameter tuning, as well as debugging.

Chapter 9: Sound Source Separation

 
This chapter include recipes for sound source separation. Similar to sound source localization, this covers the recipes in system building, parameter tuning, and debugging.

Chapter 10: Feature Extraction

 
Speech recognition requires the extraction of features from separated sound. This chapter include recipes on the introduction and extraction of features used in speech recognition It also include recipes of Missing Feature Theory used to select reliable features.

Chapter 11: Speech Recognition

 
This chapter include recipes on how to make a configuration file for Julius .

Chapter 12: Others

 
This chapter include recipes that are not found in other chapters, such as selecting a window function for frequency analysis.

Chapter 13: Advanced recipes

 
This chapter include advanced recipes, such as new functions added to HARK and connecting HARK to other systems.

Chapter 14: Sample Networks

 
This chapter include various sample network files. These samples can be used as reference in creating a network.