FFT Export is a Java-based command-line program for creating FFT analysis files from audio files. It is heavily based on Spanform, a C++ GUI program for spectral analysis that I wrote previously.
FFT Export was written in 2001 and uses the now-mostly-defunct Java Media Framework; I'm just moving it to GitHub for safe-keeping and in case anyone is interested.
java FFTExport [options]
All command-line options are optional. The options are listed below:
-afile [file]
Audio file, including full path info, to use as input. Audio can be in any format readable by the Java Media Framework. If not specified on the command-line, selection is made via a dialog box once the program starts.
-astart [seconds]
Offset (in seconds) into the sound file to begin analysis. Default is beginning of sound file.
-aend [seconds]
Time (in seconds) into the sound file to stop analysis. Default is end of sound file.
-ofile [file]
Output analysis file, including full path info. Analysis file is plain text. If the file already exists, you will be prompted whether you wish to overwrite the existing file. If not specified, selection is made via a dialog box once the program starts.
-pow2 [number]
Power of two to use for FFT analysis. Must be a power of 2 between 256 and 16384. Default is 2048. Larger numbers will create larger, more detailed analysis files, but will require more time to create. Larger numbers will also cause the analysis to 'blur' sounds together somewhat.
-cutoff [number]
Hi-frequency cutoff (in Hz). Spectral data higher than the cutoff will be ignored and not included in output analysis file. Default is 7000 and maximum is 20000.
-t [seconds]
Specifies how often (in seconds) an FFT snapshot is taken of the sound. Very small numbers will result in huge analysis files. The default value of 0.05 seconds will work well for short sounds, but might need to be adjusted for longer sounds.
-noham
Use rectangular windowing instead of the default Hamming window.
-savetemp
The program creates a temporary '.au' file during the analysis process. Choosing this option will prevent the temp file from being deleted when the program is closed.
-nonverbose
Don't write information to command line as portions of the task are completed.
-freqout
Write frequency information to analysis file. This makes the file easier to read; this information can also be inferred without writing it explicitly to the file (see below).
-help or -man
Displays information about command-line options. All other command-line options will be ignored.
-version
Displays version information. All other command-line options will be ignored.
The Analysis File is plain-text.
- Line 1 - Version Information
- Line 2 - Name of original file
- Line 3 - Information about the temporary file from which the actual analysis is taken
- Line 4 - Power of 2 used for FFT processing
- Line 5 - Hi frequency cutoff
- Line 6 - Windowing used (Rectangular or Hamming)
- Line 12 to end - Analysis Frames
Each analysis frame consists of four lines.
- Line 1 - Frame Number
- Line 2 - Two values, the maximum volume for the frame and the time of the frame. The maximum volume is the maximum raw sample value for that frame, out of 32678. This is important because frames with very low volumes contain 'static' and should probably be ignored. The time reading is simply the time of the middlemost moment of this frame, in seconds.
- Line 3 - FFT data. This line is a series of numbers separated by spaces. Each number represents the amount of spectral activity at a certain frequency. Values have been scaled to 100, so the most prominent frequency in each frame will have a value of 100. To find the frequency represented by each value, use the following formula: (samplerate[44100] * x) / powerof2 Power of 2 can be found on line 4 (see above). 'x' is 1 for the first value on the line, and increases by 1 for each value (1, 2, 3, 4, etc.) If you wish to see the frequencies directly, choose the '-freqout' option. This line then becomes many lines, with the frequencies indicated.
- Line 4 - Separating space.