details

This site is just a place for me to display my thoughts on my project and also post my web diary entries. All students are required to record our project progress in our web diaries and I thought that I might as well post them here too, just in case there are some speech enthusiasts out there.

The web diary entries are a bit formal since it’s what I have to submit every now and then to show my project’s second marker that I’m actually doing some work.

If you’re just bloghopping and somehow landed here, I suggest you take a jump to another blog right now. This site is very very very boring. You have been warned!

If the topic happens to be of interest to you or if you’re already bored anyway so you don’t mind being a little bit more bored, then you’re very welcome to read on. Please do share your speech formant ideas if you have any. The aims and description of my project are given below:

Identification and Tracking of Speech Formants in Noise

Aims

• To study the formant structure of voiced speech and methods for estimating formants and formant tracks.
• To implement and test various algorithms for estimating formants in speech signals over a wide range of noise conditions.
• If time allows, to extend the work to novel formant estimators based on existing work for detecting low level harmonic content buried in noise.

Description

• Formants describe the spectral structure of voiced speech and the estimation of formant frequencies and bandwidths is an important tool in speech analysis.
• In very bad noise, identification and tracking algorithms give inaccurate results.
• This project will study the performance of several existing algorithms in noise to assess their performance.
• Then, a new approach will be formulated and compared with the existing methods so as to gain a better understanding of performance limitations in this task.