CE.adsp: Laboratory 2

The second lab is designed to be somewhat exploratory. Follow the initial instructions that will be presented in the lab, and then Please ask one of us if you'd like some help.

If you've got this far, then you'll have already logged in and opened a browser. The lab uses a speech analysis package that was developed at UCL called SFS (Speech Filing System). During the lab, you will get to:

1. record and playback some speech, using some basic utilities;
2. plot waveforms, spectra and spectrograms for speech files and sections of them;
3. perform basic signal processing functions, such as filtering and linear prediction;
4. perform speech processing operations, such as pitch extraction, formant tracking and primitive speech synthesis.

Getting started

Connect you headphones and microphone to the appropriate socket on your workstation: usually the headphones is on the front of the box and the microphone goes in the pink socket at the back.

The first exercises will be using simple linux commands, so all you need is to begin is a teminal window, which you can select from the top menus, or from the icon on the task bar at the bottom of the screen. From the menus, also open the audio mixer and check the mixer settings for input (Mic and IGain).

Use the command rec to acquire a recording of your speech, at a sampling rate of 16 kHz. Use play to check that it has worked and that there are no clipping or noise problems. In addition to the man pages (e.g., "man rec"), there's a bit more information about these on the audio page.

Firing up SFS

The SFS software is available in a Windows emulator that can run under Linux. To start the emulator, open a terminal window and type "vm &". You'll need to log in again. When the process is complete you should see the SFS icon on your desktop.

In the remainder of this lab, you should explore the functions available within SFS. In particular, make sure you try each of the following:

1. plot waveforms, and their wide-band and narrow-band spectra for speech files and sections of them;
2. display wide-band and narrow-band spectrograms;
3. perform basic signal processing functions, such as low-pass, high-pass and band-pass filtering, resampling and linear prediction estimation of the spectral envelope;
4. perform speech processing operations, such as pitch extraction (comparing the difference methods available), formant tracking and formant speech synthesis.

• Speech and audio utilities
• SFS homepage
• SFS tutorial

If you have any other specific questions outside of the lab, you can email me, and I'll do my best to respond swiftly.