warble

A project to try to detect melody patterns in recorded voice. The eventual goal is to be able to detect sentence type (declarative, yes/no question, wh-question, tag question, echo question, command, etc.) and possibly other things.

The (informal) report I wrote for this project can be found here

Dependencies

• python (2.7)
• matplotlib (1.2.1)
• aubio (0.4.0-alpha)
  • libsndfile
  • libsamplerate
  • FFTW
  • SWIG (to build python modules)

Error building libsndfile on osx

• I ended up commenting out #include <Carbon.h> in programs/sndfile-play.c
• built fine on linux

Error building aubio:

• Download release version from website (0.3.2)
  • Add libm to linker flags: LDFLAGS=-lm make
• or download latest from repo (0.4.0-alpha)
  • Add libm to linker flags: LDFLAGS=-lm ./waf build

TODO

• add tests!
• add a real options parser
• intonation filtering based on time
• weight pitches based on duration or pitch change
• smoothing (averaging) filter