cs51 final project, harvard university nikhil benesch, joseph botros, francis loh

• Python 2.7.4
• NumPy/SciPy
• nose (Python test runner)

These instructions apply to OS X Mountain Lion.

1. Install Xcode & Xcode Command Line Tools
2. Install Homebrew. Detailed instructions.
3. brew install python
4. Ensure /usr/local/bin/ occurs first in /etc/paths/.
5. Place /usr/local/share/python on your path.
6. Run pip install -r requirements.txt.
   • You may need a fortran compiler. brew install gfortran does the trick.

In short, just get the requirements installed somehow.

###usage mootifs.py is a command-line interface to our code. From within the appropriate virtual environment, run

python mootifs.py [command] [options...]

Git should set the executable bit, so you can invoke ./mootifs.py directly as well. Run the command with no options to see brief usage instructions, reproduced below.

usage: mootifs [command]
commands:
    bpm [wav]                    - detect bpm of wav
    instrumental [in] [out]      - extract instrumentals
    transpose [in] [out] [shift] - change pitch by shift factor
    mash [file] [file] ... [out] - generate mashup from files
    wav [file]                   - find motifs from wav file
    csv [file]                   - find motifs from csv file

####bpm Takes the path to one wave file wav. Outputs the beats per minute.

####instrumental Takes the path to an input wave file in and output wave file out. Writes an instrumental version to out.

####transpose Takes the path to an input wave file in and output wave file out. Scales song by shift factor, thus changing the pitch. Use floats greater than 1 to lengthen song and lower pitch, or floats less than 1 to shorten song and increase pitch.

####mash Generate a mashup from files and write out into out file. Uses the first file passed as an instrumental backdrop, and cycles through motifs of all additional files passed in.

Note that this command can take quite a while to run, especially on larger wave files. We recommend creating mashups with no more than three songs.

####wav Find motifs within wave file. This command uses less downsampling than the mashup command to provide more resolution for the MTA. Thus, the motifs returned by this command may not be the same motifs used in mashup generation.

Motifs are output in the following format

    ['?' '?' '?' '?' '?']
        start ##s, length ##s
        start ##s, length ##s
        start ##s, length ##s

where each ? represents a symbol in the motif, and each following line denotes a discovered instance of the motif, starting at time ##s with length ##s.

####csv Find motifs with csv file. This command takes approximately 10 minutes to run on a data file with 1000 entries. The CSV should contain only one column of data, where rows are separated by newline characters '\n'.

Motifs are output in the following format

    ['?' '?' '?' '?' '?']
        start ##, length ##
        start ##, length ##
        start ##, length ##

where each ? represents a symbol in the motif, and each following line denotes a discovered instance of the motif, starting at row ## with length ##.

####general notes Not all wave file formats are supported. In particular, wave files must be uncompressed, with samples stored as 8-, 16-, or 32-bit integers. If a particular wave file is denoted unreadable, try re-exporting from audio editing software with all metadata removed.

###testing Run nosetests to invoke the nose test runner. This will run autodiscover all tests located in files in the tests/ diretory.

Testcases for algorithms are enabled by default. Testing the music and mashup modules require generation of wav files and are commented out by default. To run these tests, uncomment the tests in the appropriate files (tests/test_music.py and tests/test_mashup.py).