Back to files

Readme file for the project SAA

SAA is a tool for analyzing audio signals in terms of its content. It aims to have the following capabilities.

1. Replicate human hearing system in terms of separating sources, recognizing speech and thus analyze the auditory scene.
2. Listen to the separated sources.
3. It will accept inputs from files or a two microphone array setup whose construction would be described along with the project or else were.

These processes can be done on any segment of the speech which has been input to the program. Other than this specialized function the software would do other processing like, synthesizing spatialized sounds from mono recordings, saving, cutting, splicing, normalizing, analysis via spectrogram, filtering using various kinds of filters and re-sampling.

At its present stage of development it has the capability to

1. Record and save files, mono and stereo
2. Normalize the the signals
3. Play the processed signal
4. Separate sources using recordings from a two microphone array (It is not real time at this stage so recording has to be done as a first step and the processed)
5. View and listen to the separated sources and if necessary save them.

Program is supported on Linux. It is written in written in C and uses the following libraries.

        a)xforms.        avaialble at         http://world.std.com/~xforms/
        b)fftw3 libraries. avaialble at        http://www.fftw.org/
        c)libsnfile libraries. avaialble at   http://www.zip.com.au/~erikd/libsndfile/

Make sure they are installed on your system or compile will fail.

To INSATALL/install/Install

1 )gunzip and untar.

tar -xzv saa.tgz

2)change directory to saajoby

cd saajoby

3)Make

make

4)Log in as root and change to the saa directory. Copy saa executable and the icon to bin and pixmap directories

cp saa /usr/local/bin/
cp SAA_small.xpm /usr/share/pixmaps/

Buttons under Edit are not functional yet.

How to use?

1)Opening an existing file.

File->Open

2)To play

Signals -> Play

3)To record a stream.

1. Choose the bits (click on the checkbox)
2. Choose the sampling rate (click on the checkbox)
3. Choose the number of channels. (click on the checkbox)
4. Plug in the sound source like mirophones to the sound card.
5. Check your mixer settings (not yet controllable from saa)
6. Choose new file File -> New
7. Set the time for recording by tyong it in the box

   (Not yet controlable in between the recording)

8. Start recording Signals -> Record

4)To normalize the ampletude to 1

Process->Normalize

5)To separate the sources recorded from a stereo recording with microphones placed at 13cm appart. Use cardioid microphones. Use a preamplifier so that you can make use of the line in of your sound card for stereo recordings. (Such a kit can be sold as a product along with this software :)). It works reasonably well in a less reverberant room.

a)Record a peice with more that one source or open a previously recorded file. (At present the algorithm works only are 44.1kHz.) b)Start separation by

Process->Sep_sorcs
c)When the processing is over a window pops up from which you can choose and play the desired sources. d)You can save these sources if you want using the buttons on this new window.