Praat for Beginners:
Tutorial: Sound file formats
- Sound files can be very large, depending on time and sampling rate (e.g. speech recordings need at least 20000Hz sampling to capture [s t f], hifi recordings need higher sampling rates than 40000Hz).
- One minute of CD stereo quality (44100Hz sampling and 16-bit resolution) is around 10MB.
- One minute of mono telephone quality (8000Hz sampling, 16-bit) is just 1MB.
- The large size of sound recordings means they occupy a lot of disk space, and makes it time-consuming or impractical to send them round the web, especially by e-mail).
- Consequently ways have been devised to compress sound files to smaller sizes.
- There are two kinds of compression, lossless and lossy.
- Lossless compression does not discard any information, and is reversible, recovering all the original sound data when decompressed. Lossless compression is useful for speech analysis because it preserves the original recording intact.
- Lossy compression discards information that can never be restored and inevitably means poorer sound quality when decompressed. Worse still, every time a lossy-compressed file is edited and saved anew there is further deterioration. Lossy compression is usually achieved by degrading the lower frequencies and the higher frequencies, which usually renders the recording useless for speech research.
- Note that converting a lossy-compressed sound file to a lossless sound file, say MP3 to WAV, does not improve or restore the existing degradation of the MP3, the poorer quality MP3 file simply becomes a poorer quality WAV file. The advantage of conversion is that there is no additional degradation each time you work on your file and save it again.
- If you are nevertheless tempted to use lossy mini recorders:
- Remember that field expeditions are costly; cheap lossy recorders are not true savings.
- Some mini recorders will also do uncompressed recordings; if so, ignore the lossy settings.
- Make comparative tests before you do fieldwork, to make sure that the specific speech characteristics you intend to measure have not been degraded.
- Log all lossy-compressed recordings and subsequent converted sound files as having been compressed at some time, so that no-one will think they are full quality. Better still, include and keep the method in the filenames you choose. A filename like dagestaniMP3.wav will always remind you how the degraded recording was originally acquired.
2. Lossless sound formats
Praat handles the following uncompressed and lossless-compressed sound file formats
- AIFF: For Macintosh.
- AIFC: A variant of the AIFF file format for Macintosh that is capable of compression. Praat reads and saves uncompressed AIFC files, but does not support compressed AIFC files.
- Binary: a standard format for sound files, independent of computer system (if nothing else works, this could be a way of exporting data to other programs on other systems)
- FLAC: An open source lossless compression format.
- Kay: The format used by the Computer Speech Laboratory (Kay Corporation)
- NeXT/Sun: A format used by NeXT computers and Sun computers
- NIST: A format used by theTIMIT data base (American English, produced by Texas Instruments and Massachussetts Institute of Technology)
- WAV: A format used by MS Windows
Your choice of format is mainly determined by what is recognised by your computer and its sound system, by any other computer systems you intend to move between, and by any external program you intend to use your data with.
As long as you are only using Praat you are free to use at least AIFF, Binary, NeXT, NIST, or WAV, since the program can read them all independently of any computer system you may happen to move between.
3. Lossy formats
- Atrac: a proprietary sound data compression used in mini recorders and mini players (exclusive to some Sony products). Praat does not handle the Atrac format, so these files will need to be converted to one of the lossless formats listed above. For Atrac, you probably need the software supplied with the device. More recently, Sony have developed a lossless variant of Atrac, but this is not supported by Praat either.
- MP3: is a patented lossy method, widely used in mini recorders and mini players. Praat opens MP3 files but does not save them. You might find conversion to other sound formats is available in some commercial or shareware sound programs (or your computer sound system might also come with software that converts MP3, check your manual).
- MP3: See also MP3 and recording.